fix(test): bump L1→L2 wait in e2e_p2p_add_rollup to 300s

[AztecBot]

The merge-train/barretenberg train PR (#23418, head ae4f1646f7) failed CI on e2e_p2p/add_rollup.test.ts with:

TimeoutError: Timeout awaiting L1 to L2 message ... ready
  at makeMessageConsumable (add_rollup.test.ts:322)
  at bridging              (add_rollup.test.ts:335)
  at Object.<anonymous>    (add_rollup.test.ts:610)

The failure was inside the second bridging() call — the one targeting the freshly-registered new rollup, executed right after the test warps the L1 clock forward by ~500 epochs. Validator logs show the chain repeatedly attempted to land a checkpoint, hit Contract class ... already exists, cannot add again at block 6 on store-updater replay, and pruned back to block 15. The L1→L2 message never reached the inboxLag: 2 checkpoint the test polls for, and the inner 120 s budget expired.

The two commits new to the train in this run (feat(ci): refresh Chonk inputs (#23400) and fix(ci): harden Chonk refresh post-action (#23419)) touch only CI scripts and barretenberg/ts/ CRS initialization. Nothing in archiver/validator/sequencer/p2p changes, so this is a pre-existing flake on the test, not a regression introduced by the train.

This raises the inner wait from 120 s → 300 s. The outer per-test TIMEOUT=20m in yarn-project/end-to-end/bootstrap.sh and the jest.setTimeout(20 * 60 * 1000) in the file both stay — neither is hit by this bump.

Full diagnosis (validator log excerpts, archiver throw site, ruled-out alternatives): https://gist.github.com/AztecBot/2006827080c810656b90ef3e5c8bd5e7

Failing run: https://github.com/AztecProtocol/aztec-packages/actions/runs/26132528825
Failing log: http://ci.aztec-labs.com/1779234898530459

ClaudeBox log: https://claudebox.work/s/27f13a9bd0fcf5e1?run=1