perf(e2e): batch same-sender setup txs into single txs

[spalladino]

Context

#24569 overlapped the fees harness's independent setup txs with Promise.all, but the fix-phase capture showed only ~1/3 of the projected gain (−162s). The overlap was real (tx:mint busy/total ≈ 0.68), yet concurrent txs still land in consecutive 12s slots: the PXE serializes simulateTx/proveTx on a single queue, so the second concurrent tx is only proven after the first and misses the first's slot. Overlapping cannot remove whole slots — only batching can.

This PR batches same-sender setup calls into a single BatchCall tx: one proof, one tx, one slot.

What was batched

FeesTest.applyFundAliceWithBananas (end-to-end/src/single-node/fees/fees_test.ts) previously sent alice's private and public banana mints as two concurrent txs. Both are alice sends on BananaCoin touching disjoint balances, so they now ride one BatchCall([mint_to_private, mint_to_public]) tx. The private-balance before/after assertion is preserved as the behavioral check; the public mint's effect is checked by the downstream tests that spend alice's public bananas. Precedent for the pattern already lives in this tree (mintNotes in fixtures/token_utils.ts batches multiple mint_to_private calls).

This touches the fees suites that fund alice in setup: account_init, gas_estimation, private_payments, and fee_juice_payments.

Surveyed and skipped

• applyFPCSetup: a single deploy:fpc tx — no same-sender pair inside the phase. Folding the FPC deploy into the mint batch (one class publication + app calls is within the class-log limit) would remove the remaining consecutive slot, but it merges two harness phases that four test files compose differently via Promise.all, and deploy-inside-BatchCall is not a proven pattern here — a restructure, not a mechanical merge. Left as a possible follow-up.
• applySponsoredFPCSetup: no on-chain tx — it only calls wallet.registerContract locally; the sponsored FPC is genesis-funded and its class preloaded. Nothing to batch.
• applyBaseSetup chain: the auth-registry publish short-circuits (genesis-seeded by perf(e2e): seed standard contracts at genesis #24568) and applyDeployBananaToken is a single deploy tx. No consecutive same-sender pair.
• Two different contract deploys cannot share a tx (MAX_CONTRACT_CLASS_LOGS_PER_TX = 1), which rules out any two-deploy batch (e.g. cross-chain token+bridge).
• Cross-chain harness deploys and bot/src/factory.ts are out of scope (the latter lives on perf(bot): parallelize independent bot factory setup steps #24581's stack).

Local span evidence (account_init, 5/5 green on all runs)

Suite-scoped beforeAll spans from TEST_TIMING_FILE captures:

run	tx:mint count	tx:mint busyMs	deploy:fpc busyMs	beforeHooksMs
baseline (before)	2	23618	23713	71607
run 1 (after)	1	11834	24065	72372
run 2 (after)	1	11807	23743	71678

The mint phase drops from two consecutive slots (~23.6s) to one (~11.8s), one proof and one tx fewer per shard. account_init's total beforeAll is flat locally because its mint phase runs concurrently with applyFPCSetup and deploy:fpc still pays 2 slots (PXE proves the mint batch first), remaining the phase's critical path — the direct wall-clock win lands on suites where the mints are the phase (e.g. fee_juice_payments), and the fewer-proofs contention win applies everywhere.

Stacking

Stacked on #24569 (spl/e2e-overlap-setup-txs), itself stacked on #24564. Retarget to merge-train/spartan-v5 once #24569 merges.

Fixes A-1409

Measured impact

The batch lands exactly as the local runs predicted: 13 mint txs merged run-wide, with the wall-clock
win concentrated on the suites where the mint phase is the critical path.

• tx:mint: 42 → 29 occurrences (one merged per fees shard); busyMs 523s → 431s (−92s of tx/proving
  work). In the PR run busyMs equals totalMs — there are no overlapping mint txs left to union.
• Suites where the mint is critical-path: failures −12.0s, fee_juice_payments −11.5s (≈ one 12s
  slot each, as designed).
• Suites where the mint overlaps applyFPCSetup are flat as expected: account_init −0.4s,
  private_payments.parallel +2s (×8 shards), gas_estimation.parallel +3s (×3) — deploy:fpc is
  unchanged (13 occurrences, ~23s / 2 slots each) and remains their setup critical path, so removing
  a slot from the mint branch of the Promise.all doesn't move the union.
• Run-wide sum over common suites: −8s, i.e. neutral outside the two critical-path suites; the noise
  envelope on untouched suites in this run pair is ±20–37s (bot −20s, private_initialization +37s).

Net: ~−23s wall on two suites plus 13 fewer txs/proofs per run (less proving contention and one less
tx per fees shard through the sequencer). The remaining ~12s/shard lever on the overlapped fees
suites is folding the FPC deploy into the same batch (one class publication + app calls is within
MAX_CONTRACT_CLASS_LOGS_PER_TX), noted above as a follow-up.

Baseline CI run 1783394912998666 (#24569 ci/x-fast, the stack base). PR CI run 1783432285938536
(ci/x-fast).

[spalladino]

Folded into #24569 per review: the two PRs touched the same mint path in fees_test.ts (this one rewrote the Promise.all overlap that #24569 introduced into a single BatchCall), so they are now a single PR. The commit here (7b9cd12) was fast-forwarded onto spl/e2e-overlap-setup-txs unchanged, and #24569's description now covers both changes, including this PR's measured impact. Fixes A-1409 moves there too.