Add SimRequest::ValidateMerged and dispatch merged blocks through the simulator

[0w3n-d]

Issue: #500 (step 2 of 4)

Base branch: od/merged_block_sim (step 1, #502) -- stacked per CONTRIBUTING.md, not yet merged. Diff will shrink once #502 merges and this retargets to develop.

What this PR does

Adds SimRequest::ValidateMerged/SimResult::ValidateMerged so a merged block from the merge builder can be simulated through the existing SimulatorTile machinery. A merged block is converted into a synthetic SignedBidSubmission (BidTrace.block_hash/parent_hash/gas_limit/gas_used derived from the execution payload, value from proposer_value, proposer_fee_recipient from the request, builder_pubkey/proposer_pubkey/signature zeroed -- the simulator never checks the BLS signature and these fields are otherwise cosmetic), so it reuses the same SSZ/JSON dispatch already used for ordinary bid submissions. A new PendingMergeRequests queue (keyed by base_block_hash, since there's exactly one merge builder connection rather than many builder pubkeys) gives merged-block requests the same load-shedding behavior under simulator capacity pressure as ordinary submissions get from PendingRequests.

SimulatorTile gains read access to merged_blocks (main.rs reordered so it's constructed before SimulatorTile::create instead of after). The auctioneer's FromSimMsg consumer, which previously did an irrefutable let SimResult::Validate(..) = .., now ignores SimResult::ValidateMerged results with an else { return } -- it doesn't act on them in this step.

What this PR deliberately does not do

Nothing sends a ValidateMerged request yet, and nothing consumes SimResult::ValidateMerged. BlockMergingTile isn't wired up to dispatch requests or read results back, and there's no builder-attribution/disable-switch/Discord-alert logic. That's step 3.

Tests

Written first, reviewed, then implemented against. A live end-to-end HTTP dispatch test isn't feasible here: spawn_sim/spawn_merge_sim dispatch via spawn_tracked!, which panics unless a process-global RUNTIME was already set up via init_runtime(&RelayConfig) -- a one-time, real-core-pinning init meant for actual process startup, and spawn_sim itself has no existing test either. So tests target the new, genuinely error-prone synchronous logic:

• merged_block_to_submission_derives_bid_trace_from_payload_and_context / merged_block_to_submission_converts_appended_blobs_to_blobs_bundle: the merged-block -> synthetic-submission conversion.
• pending_merge_requests_evicts_older_same_base_block / pending_merge_requests_keeps_existing_if_new_is_older / pending_merge_requests_next_req_returns_and_removes / pending_merge_requests_clear_empties_queue: the new queue's eviction/ordering.

just fmt-check, cargo clippy --all-features --no-deps -- -D warnings, and just test (full workspace) all pass.

Reviewer checklist

• CI (lint, unit-test) is green
• Matches the linked issue/step
• No unexplained scope creep or unrelated files touched