feat(test-benchmark): strengthen and consolidate TSTORE benchmark

[chfast]

Description

Rework test_tstore into a single parametrized worst-case benchmark: key = GAS (unique slot per op, no memory counter) or PUSH0 (fixed slot), value = COINBASE (a nonzero op) so each write actually inserts an entry. A zero value lets clients skip the insert and silently hide store growth, so the test asserts the block coinbase is nonzero.

The tload parameter adds a transient read (TLOAD and TSTORE share the same 100 gas cost):

• none: TSTORE only.
• hit: seed the value once, store it under the key and read it back, feeding the result into the next TSTORE so the read is used, not discarded.
• miss: store the key, then read a fresh, never-written key so TLOAD misses on the growing store; the value stays COINBASE (nonzero).

This subsumes the bloatnet test_tstore_same_key (fixed key) and test_tstore_unique_keys (unique key), which are removed along with their now-empty module.

Related Issues or PRs

N/A.

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.

[chfast]

Benchmark results — geth & besu

Filled fixture executed at Osaka, 300M gas via block-test
(evm blocktest / evmtool block-test), single machine, best-of-3 (geth) /
best-of-2 (besu).

Clients: geth evm 1.17.2-stable, besu evmtool 26.8-develop.

key	tload	geth MGas/s	besu MGas/s
unique (GAS)	none	141	39
unique (GAS)	hit	270	56
unique (GAS)	miss	288	56
fixed (PUSH0)	none	1222	385
fixed (PUSH0)	hit	1231	462
fixed (PUSH0)	miss	1254	455

The unique (distinct transient slot per op) variants are ~9x slower than the
fixed baseline — the intended worst case. tload=hit/miss add a TLOAD, so
fewer TSTOREs run per unit gas and the per-tx transient store grows less,
giving higher throughput than tload=none.

Notes:

• Single-machine, relative numbers (not hardware-normalized); besu is a locally
  built dev evmtool, so treat its absolutes as indicative.
• Amsterdam was tried first but no client runner supports it yet (geth's t8n /
  blocktest excludes it), so these are on Osaka.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.77%. Comparing base (786c0d4) to head (123ddc8).
⚠️ Report is 1 commits behind head on forks/amsterdam.

Additional details and impacted files

@@                 Coverage Diff                 @@
##           forks/amsterdam    #3441      +/-   ##
===================================================
+ Coverage            93.54%   93.77%   +0.22%     
===================================================
  Files                  624      624              
  Lines                37074    36905     -169     
  Branches              3394     3326      -68     
===================================================
- Hits                 34681    34606      -75     
+ Misses                1644     1597      -47     
+ Partials               749      702      -47     

Flag	Coverage Δ
unittests	93.77% <ø> (+0.22%)	⬆️

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.