fix: transfer() rejected a fee exactly equal to transferMaxFee, unlike sendTransaction()

[ayushsingh82]

Summary

• transfer()'s max-fee check used prepared.fee >= transferMaxFee, rejecting a fee exactly equal to the configured cap.
• sendTransaction()/signTransaction() use > transactionMaxFee, allowing a fee exactly equal to the cap.
• git blame shows the > boundary was a deliberate fix applied to transactionMaxFee (commit "updated max fee checks, testing") that was never mirrored to transferMaxFee, which was left behind with the original >=.
• Concretely: call quoteTransfer(), pass its returned fee straight back in as transferMaxFee, and transfer() throws 'Exceeded maximum fee cost for transfer operation.' even though the fee didn't exceed the cap — it equaled it. The identical scenario through sendTransaction()/transactionMaxFee succeeds.

Change

• src/wallet-account-evm-erc-4337.js: transfer()'s check changed from >= to >, matching sendTransaction()/signTransaction().
• tests/wallet-account-evm-erc-4337.test.js: updated the existing test that had codified the old (inconsistent) behavior — 'should throw if the fee equals the transfer max fee configuration' — to assert the corrected, consistent behavior instead: 'should allow a fee exactly equal to the transfer max fee configuration', mirroring the equivalent sendTransaction test already in the suite.

Breaking change note

Per repo convention, flagging explicitly: this changes observable behavior of transfer() — a transfer whose fee exactly equals transferMaxFee will now succeed instead of throwing. No other public API surface changes.

Test plan

• npm run lint (standard) passes
• npm test passes — 92/92 unit tests (3 suites)
• npm run build:types passes (generated .d.ts diff not included — it's drift from a different local tsc version across unrelated files, not caused by this change)
• Confirmed the updated test previously passed against the old >= implementation (i.e. the old behavior was real, not a flaky test) before flipping it to assert the fixed behavior

[ayushsingh82]

@jonathunne could you take a look when you have a chance? Thanks!

Summary of the fix: `transfer()`'s max-fee check used `prepared.fee >= transferMaxFee`, rejecting a fee exactly equal to the configured cap — while `sendTransaction()`/`signTransaction()` use `> transactionMaxFee`, allowing an exactly-equal fee. Git blame shows the `>` boundary was a deliberate fix applied to `transactionMaxFee` (commit "updated max fee checks, testing") that was never mirrored to `transferMaxFee`. Fixed `transfer()`'s check to `>` to match, and updated the existing test that had codified the old behavior. Flagging as a behavior change per repo convention: a transfer whose fee exactly equals `transferMaxFee` now succeeds instead of throwing.