misc: use fvm-anvil to run end-to-end performance benchmarks in CI

fvm-anvil allows us to measure the actual gas usage of our contracts.
I'm in the process of pushing a FWSS benchmark harness to fvm-anvil.
We could use it to measure the impact of our changes on FWSS gas usage in CI.
I'm not sure if it will be fast enough to do it on a per-commit basis.
Maybe a more limited set of benchmarks will be fast enough to run per commit, since the current set spans a parameter matrix (about 100 different runs).

Start by reviewing the FWSS benchmark harness being pushed to fvm-anvil and the current CI setup. Determine whether the roughly 100-run parameter matrix can run per commit or whether a smaller benchmark set is needed. Done means CI measures and reports FWSS gas usage for an agreed benchmark scope without making per-commit runs too slow.

Written by the indexing model from the issue text.