build(deps): bump chain33 to v1.72.4

[vipwzw]

Bumps github.com/33cn/chain33 from v1.72.0 to v1.72.4, picking up the 1.72.1–1.72.4 upstream fixes.

What comes in

• blockchain — the rollback floor is now bounded by the archiver's frontier, and a rollback target below the archived block bodies (or one the node could not start from) is refused.
• address — CheckAddress walks the drivers in a fixed id order, so the reported error no longer depends on Go's randomized map iteration; the address cache is keyed by the enabled drivers.
• system/dapp — CheckAddress tolerates the legacy checksum error below ForkBase58AddressCheck. The deterministic driver order makes a 25-byte address report ErrCheckChecksum where it could previously come back as ErrCheckVersion, so without this gate such an address would flip from "sometimes accepted" to "always rejected" on already-executed blocks.
• account — GenesisInit rejects a negative amount rather than persisting a negative balance; the upper bound stays at MaxTokenBalance so existing blocks remain replayable.
• db — bumps goleveldb past the MANIFEST rotation table-file leak (transitively raises our indirect github.com/syndtr/goleveldb pin) and caps the leveldb write buffer.
• p2p — verLimit no longer blacklists peers on a higher minor version.

Scope

Dependency-only change plus the transitive goleveldb pin. No new forks are registered anywhere in this range, so the fork toml templates need no new entries (unlike the v1.72.0 bump, which added ForkAccountBlacklistV2).

Verification

• go build ./... — clean apart from plugin/dapp/cross2eth/contracts/gnosis/{bsctest,chain33test}, which fail identically on master (pre-existing generated packages with no main).
• go test -parallel=8 $(go list ./... | grep -v pbft) — 124 packages ok, 0 failures.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 29.12%. Comparing base (eaf1546) to head (1b86e4b).
⚠️ Report is 35 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1324      +/-   ##
==========================================
- Coverage   29.12%   29.12%   -0.01%     
==========================================
  Files         382      382              
  Lines       73475    73477       +2     
==========================================
- Hits        21400    21399       -1     
  Misses      50087    50087              
- Partials     1988     1991       +3     

☔ 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[vipwzw]

Note on the red ci_rgbx check — it is not caused by this bump.

The job times out at wait auto DKG commit by neutrino+tss (a 120s poll for a non-empty tssAddress). I reran it once, and it failed identically. To separate the bump from the environment I then reran the last green ci_rgbx on master — run 35298147251 at 1b81b6417, i.e. this branch's own base commit with chain33 still at v1.72.0. It fails today with the same error:

[2026-10-01 05:06:16] wait auto DKG commit by neutrino+tss
[2026-10-01 05:08:40] ERROR: auto DKG commit timeout

So the step is currently broken on master independent of this PR. Everything else is green, including unit-test, coverage, check_fmt, ci_evm, ci_cross2eth, ci_relay, build on windows and Test GOARCH 386.