refactor(ci): run geth benchmark CI only on benchmarks/** branches

[marioevz]

Description

Benchmark CI previously coupled an external, spec-matched geth dependency into checks that ran on (nearly) every PR: benchmark.yaml's bench-gas job fills fixtures with geth and re-verifies them against EELS, so whenever the EELS spec moved ahead of the pinned geth ref, unrelated spec PRs failed the benchmark job (see #3214: EELS on devnet-8 vs geth on glamsterdam-devnet-7).

This PR separates benchmarking so that mainline development is never blocked by an external client, while the geth-backed work moves to dedicated benchmarks/** branches:

• test_benchmarking.py defaults to EELS. The benchmark-plugin fills previously hardcoded --evm-bin (defaulting to geth's evm). They now omit --evm-bin when EVM_BIN is unset, falling through to fill's in-repo EELS t8n; setting EVM_BIN still targets a specific binary. This removes the only reason those tests needed an external client.
• Plugin unit tests fold into the normal unit-test suite. The --ignore=.../test_benchmarking.py is removed from just test-tests, so the benchmark-plugin tests now run as part of the standard unit tests on every forks/** PR via test.yaml — on EELS, with no geth. The redundant test-tests-bench recipe is deleted.
• test.yaml behaves differently on benchmarks/**. On those branches it builds the pinned geth and sets EVM_BIN, so the same benchmark tests run against a real client; on every other branch the build step is skipped and the tests use EELS. (EVM_BIN is read only by test_benchmarking.py, so this has no effect on the rest of the suite.)
• benchmark.yaml runs only on benchmarks/**. It now holds the geth-backed sanity checks (bench-gas cross-check, bench-opcode, bench-opcode-config) and the fixture-artifact build. These branches pin a spec version together with a matching geth ref, so the geth↔EELS cross-check is expected to agree there.

Net effect: forks/** PRs get full benchmark-plugin coverage on EELS with no external dependency, and the geth cross-check/fills run where geth and the spec are deliberately kept in sync. This is the first step of a larger benchmarking separation; creating the benchmarks/amsterdam branch and the release/publish flow are follow-ups.

Files changed:

• packages/testing/.../filler/tests/test_benchmarking.py — EELS-by-default _evm_bin_args() helper.
• .github/workflows/test.yaml — add benchmarks/**; conditional geth build + EVM_BIN in test-tests.
• .github/workflows/benchmark.yaml — scope to benchmarks/**; drop unit-tests (now covered by test.yaml).
• Justfile — un-ignore the benchmark tests in test-tests; delete test-tests-bench.
• docs/getting_started/verifying_changes.md — update the benchmark row.

Related Issues or PRs

#3244

Checklist

• Ran fast static checks to avoid CI fails, see Code Standards & Verifying Changes: just static
• PR title has the form <type>(<area>): <title>, where <type> and <area> come from an appropriate C-<type>, respectively A-<area>, label. The title should match the target squash commit message.

Cute Animal Picture

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.41%. Comparing base (c69d54b) to head (d1c4be6).
⚠️ Report is 4 commits behind head on forks/amsterdam.

Additional details and impacted files

@@               Coverage Diff                @@
##           forks/amsterdam    #3249   +/-   ##
================================================
  Coverage            93.41%   93.41%           
================================================
  Files                  625      625           
  Lines                37035    37035           
  Branches              3386     3386           
================================================
  Hits                 34597    34597           
  Misses                1668     1668           
  Partials               770      770           

Flag	Coverage Δ
unittests	93.41% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[spencer-tb]

LGTM!

cc @danceratopz for when you are back