BitBox02 firmware

maintenance: Rename "callback param" to "user data"

"user data" is the conventional term for the user provided pointer that
is returned in a callback.

assertions: Add nickez assertions for v9.23.2

releases: add sutterseba v9.23.2 assertion

releases: add v9.23.2

sd.c: always pause clock after sdcard check

When the sdcard is inserted, the sdcard check runs into the retry
loop. In that case, the clock was not paused, which could lead to
touch issues.

rust: change imports to util::bb02_async

rust: move async helper functions to the util crate

So the bitbox02 crate can also use it. It will be needed to be able to
use block_on() in the tests there.

bitcoin: add support for OP_RETURN outputs

We enforce 0 value on them, so the confirmation screen is not a the
usual recipient component, but can be a full screen confirmation not
showing the amount. We use verify_message flow as it handles both
ascii/binary.

We also only support OP_RETURN outputs with one data push, though an
OP_RETURN output could contain multiple data pushes. This restriction
is for simplicity and because we don't know of a use case. In the
future, support for this can be added if needed.

bitcoin/common: use bitcoin crate to compute pkscript

Remove our own code to improve clarity.

bitcoin/common: add unit test for pk_script()

bb02_async: compile block_on only in testing

Not needed anymore in production code. As a result, we can also remove
the screen processing, which is not relevant to the Rust unit tests.

keystore: move unlock_bip39() to Rust

With the goal of making it async.

keystore: split keystore_unlock_bip39() into multiple functions

Simpler to migrate it to Rust, with the goal of making it async.

test: apply clang-format

rust: update to toolchain 1.90.0

This saves ~2-3kB in binary space in the multi firmware.

`clippy::collapsible-if` is disabled because I didn't like some of the
collapse suggestions. Readability seems worse rather than better.

Applied this to remove duplicate deps, choosing the higher of version
each time.
```
cargo update --precise 0.1.74 --package libc
cargo update --precise 1.0.1 --package cfg-if
```

Rust vendoring now for some reason adds Cargo.toml.orig files, which
part part of the crate checksum, so they are removed from the
.gitignore and committed.

maintenance: upload simulator artifact in CI.

This way we can iteratively use new versions of the simulator without
the need for a new release.

We remove the old step that just built the simulator, but did nothing
with it, and include the simulator in the process that builds and
uploads artifact for several targets.

chore: fix warnings.

Fix warnings due to either unsigned/signed type mismatch in comparison,
or ignored return values that should have been checked.

py: bump version and add CHANGELOG entry for btc_xpubs

build: fix sanitizers

The recommended way to enable sanitizers nowadays is to add
`-fsanitize=*` as a compiler and linking flags.

Also fix some warnings and remove some targets from "all" to speed up
builds.

Bumping macosx version to 11 since that seems to be necessary for asan
to work.

Add CI check to verify that address santizier is turned on for unit
tests

test: add option for the simulator to write to file.

If the env variable FAKE_MEMORY_FILEPATH is provided, the simulator will
write to file using three different files:

- ${FAKE_MEMORY_FILEPATH}_shared
- ${FAKE_MEMORY_FILEPATH}_app
- ${FAKE_MEMORY_FILEPATH}_eeprom

mimicking the behaviour used when reading/writing to memory.

This allows us to re-launch the simulator after making changes in the
BBApp (e.g. renaming the device, accounts etc) while testing, simulating
the effect that unplugging/replugging a device would have.

test: add option for the simulator to write to file.

If the env variable FAKE_MEMORY_FILEPATH is provided, the simulator will
write to file using three different files:

- ${FAKE_MEMORY_FILEPATH}_shared
- ${FAKE_MEMORY_FILEPATH}_app
- ${FAKE_MEMORY_FILEPATH}_eeprom

mimicking the behaviour used when reading/writing to memory.

This allows us to re-launch the simulator after making changes in the
BBApp (e.g. renaming the device, accounts etc) while testing, simulating
the effect that unplugging/replugging a device would have.

keystore: reduce SC operations in the seed check

verify_seed() after storing a seed currently re-stretches the
password, which needs 5 securechip events.

We can skip that because we still have access to the stretched
password.

If the check succeeds, but there was an error involving the password
or secure chip (e.g. the stretching result was corrupted the first
time), the consequence is that the user won't be able to unlock, but
the seed itself on the MCU, on the backup and retained (encrypted) in
seed is unaffected. That trade-off is good, because a failure
there (securechip returns no error but actually had a corruption)
should be exceedingly rare.

This effort is part of mitigating Optiga's throttling mechanism that
kicks in after 133 events - users can run into this by repeatedly
resetting/restoring).

keystore: pass seed to unlock_bip39() to reduce secure chip events

When restoring a wallet, the seed is already known, so no need to do
another `copy_seed()` which is a secure chip security event. This
reduces the number of secure chip operations when restoring.

The hashed seed is retained so it can be compared without storing it
in plaintext.

This effort is part of mitigating Optiga's throttling mechanism that
kicks in after 133 events - users can run into this by repeatedly
resetting/restoring).

reduce number of secure chip security events when creating/restoring

Before, `keystore_encrypt_and_store_seed()` (called when
creating/restoring a seed) would always be followed by
`keystore_unlock(<password>)` with the password the user just chose,
so unlock could never fail. The unlocking part costs many secure chip
operations (for stretching the password).

By making the first function already unlock the keystore, we can avoid
calling `keystore_unlock()`, reducing the number of secure chip events
by 5.

This effort is part of mitigating Optiga's throttling mechanism that
kicks in after 133 events - users can run into this by repeatedly
resetting/restoring).

keystore: move funcs

To be below the static function `_retain_seed()` that will be needed later.