[SharovBot] fix(commitment): count first-nibble branches in CanDoConcurrentNext to fix flaky test

[Giulio2002]

[SharovBot]

Problem

Test_Trie_CorrectSwitchForConcurrentAndSequential fails intermittently on CI:

--- FAIL: Test_Trie_CorrectSwitchForConcurrentAndSequential (0.00s)
    hex_patricia_hashed_test.go:280:
        Error: Should be true
        Messages: should be able to parallelize next run

Failing job: https://github.com/erigontech/erigon/actions/runs/22561824327/job/65349926766

Root Cause

CanDoConcurrentNext() only checks if nibble 0 has a branch node:

zeroPrefixBranch, _, err := p.root.ctx.Branch(HexNibblesToCompactBytes([]byte{0}))
if len(zeroPrefixBranch) > 4 {
    return true, nil
}

The test generates 150 keys using a globally shared rand.New(rand.NewSource(42)) protected by a mutex. Because multiple tests run in parallel they consume the random stream at different offsets each run. Occasionally, the resulting 150 keys hash (keccak256) such that zero or one keys have first nibble 0 — the nibble-0 branch is empty, so the function returns false even though the trie is perfectly wide enough to parallelize.

P(zero keys at nibble 0 with 150 random keccak hashes) ≈ (15/16)^150 ≈ 5.6×10⁻⁵ per run. Invisible in dev, but observable across thousands of CI runs.

Fix

Instead of probing a single arbitrary nibble, count how many of the 16 first-nibble sub-tries have stored branch data. Return true only if at least 4 nibbles are populated.

Scenario	Nibbles filled	Result
150 random keys	~16/16	true ✓
2 remaining keys (post-deletion)	≤2/16	false ✓

The threshold of 4 makes the decision independent of any particular nibble's occupancy and preserves the correct semantic: small key sets are processed sequentially.

Testing

• go test ./execution/commitment/... -run Test_Trie_CorrectSwitchForConcurrentAndSequential -count=5 → 5/5 PASS
• Full ./execution/commitment/... suite → all pass