Update to Brussels 0.6.0 (#2582)

This PR updates the release flow to pull in oxidecomputer/brussels#21.
This should be the last release process change needed before we can
integrate Brussels with the omicron releng tooling!

This also changes the configuration to match the current behavior of
sprot-release: sidecar and observer images are not included in the CoRIM
measurements anymore, and reverso and observer images will not be pulled
into omicron.

Remove old dead code (#2568)

Removed an old dead code comment I saw while looking at something else

Update versioning scheme (#2562)

This PR updates Brussels and the release workflow to support the
proposed versioning scheme in [RFD
705](https://rfd.shared.oxide.computer/rfd/0705) (tl;dr moving the
version number to the minor number rather than the patch, to support
backports). The new "Custom version number" field in the release
workflow will allow both to migrate to the new versioning scheme, and to
issue patch releases (where we'll want to bump the patch number rather
than let Brussels bump the minor number).

girl forgot you need to smuggle outputs

observer: add observer-a to SP release (#2559)

Add observer-a to brussels config so it's included in releases.

Tiny build-time optimization for config handling (#2526)

This change reworks build_notifications from re-loading the task config
from an environment variable, and then parsing as toml, once for every
task. Instead, we now just get and parse the config once, and grab what
we want from the pre-parsed table.

This was noticed incidentally while debugging #2524

cosmo_seq: mask STATUS_CML bit 1 to not get annoying pmbus alerts abo…

…ut it (#2520)

It turns out that PMBus devices like to set this bit bascially any time
they saw any kind of I2C behavior that makes them feel uncomfortable. We
probably don't want to produce an ereport every time an I2C Thing makes
a VRM uncomfortable. So this commit masks out PMBus alerts for
STATUS_CML's "other communication fault" bit, since it's annoying and
not a big deal.

This branch does a bit of extra refactoring on top of the change from
#2511. Since we are now setting even more `SMBALERT_MASK`s in
`initialize_pmbus_alerts`, I thought it would be nice to have some
shared utilities for trying to do this.

FPGA Updates to fix i2c proxy bugs

Set VLAN FID in configuration (#2501)

The MAC tables in the VSC7448 contain a map from `(source mac, filtering
id)` tuples to port. The filtering id field ("FID") allows for
VLAN-aware learning.

Unfortunately, you have to explicitly set a flag to copy the classified
VID to the FID.

If you're in a situation where you see the same MAC address on multiple
ports (which are on different VLANs), properly setting the FID is
load-bearing; otherwise, that MAC address will randomly wander between
those ports, and packets directed to that MAC address have unpredictable
deliverability.

Fixes oxidecomputer/mfg#224

Add MGS command to clear APOB (#2456)