test: mark e2e_partial_notes as flaky under pipelining

[AztecBot]

CI on merge-train/spartan failed on commit fe4792ba22 (PR #23329) with e2e_partial_notes failing as Transaction 0x1e7fd673… was dropped. Reason: Tx dropped by P2P node. CI3 log: http://ci.aztec-labs.com/1779100391339924

The failure is a pre-existing race in the pipelined sequencer's waitForValidParentCheckpointOnL1 check: when the L1 propose tx for the parent checkpoint hasn't mined yet by the time the pipelined child's parent-landed check runs, the pipelined work is discarded, the next slot has nothing to publish (propose_action_not_successful), and pending txs are dropped during handlePrunedBlocks. This is unrelated to anything in PR #23329 — its only non-test diff is an error-handling tweak inside simulatePublicCalls (the RPC simulate path).

Adding a narrow flake entry keyed on the Tx dropped by P2P node error so the train can advance, with @santiagopalladino as owner since he owns the pipelining work.

Full analysis (timeline, log refs, comparison with surrounding train commits, root-cause notes for the publish-race): https://gist.github.com/AztecBot/431b9a3fdeb7e23797fa589f7f6a3431

Recommend adding ci-skip and force-merging via the bypass-merge button so the merge train can move past fe4792ba22.

ClaudeBox log: https://claudebox.work/s/78410c5e31df03fc?run=1

[AztecBot]

Automatically closing this stale claudebox draft PR (no updates for 5+ days). Re-open if still needed.