[qs] dedup requests to batch fetcher and simplify payload manager

[ibalajiarun]

Description

This PR aims to simplify the PayloadManager logic to retrieve batch payload from quorum store. To reduce complexity, we no longer cache transactions in PayloadManager. As long as quorum store stores the payload in memory, retrieval should be very fast and does not require caching in payload manager.

• Splits payload_manager.rs into separate modules one each for direct mempool, consensus observer, and quorum store.
• QuorumStorePayloadManager always requests QuorumStore for payload which will return a future that resolves into the payload upon completion. On prefetch, this future is ignored and BatchReaderImpl will spawn a task to drive the future.
• BatchReaderImpl implements deduplication logic to ensure there is only one inflight fetch request per batch. It returns a shared future that the caller can await on to get the payload. The future is driven by a tokio task, so the caller don't need to .await immediately to drive the future.

TODO: add counters to measure retrieval duration

[trunk-io]

⏱️ 36m total CI duration on this PR

Job	Cumulative Duration	Recent Runs
test-target-determinator	16m	🟩 🟩 🟩
check-dynamic-deps	9m	🟩 🟩 🟩 🟩
rust-cargo-deny	5m	🟩 🟩 🟩
semgrep/ci	2m	🟩 🟩 🟩 🟩
general-lints	1m	🟩 🟩 🟩
file_change_determinator	45s	🟩 🟩 🟩 🟩
file_change_determinator	37s	🟩 🟩 🟩
permission-check	16s	🟩 🟩 🟩 🟩 🟩
permission-check	13s	🟩 🟩 🟩 🟩 🟩
permission-check	7s	🟩 🟩 🟩
determine-docker-build-metadata	6s	🟩 🟩 🟩

_{settings ⋅ feedback ⋅ docs ⋅ learn more about trunk.io}

[github-actions]

✅ Forge suite realistic_env_max_load success on 580c7946be44a8f68e99b2968f9bfc93f58ad767

two traffics test: inner traffic : committed: 13497.42 txn/s, submitted: 13498.26 txn/s, expired: 0.84 txn/s, latency: 2940.67 ms, (p50: 3000 ms, p70: 3000, p90: 3200 ms, p99: 3900 ms), latency samples: 5131980
two traffics test : committed: 99.97 txn/s, latency: 1386.83 ms, (p50: 1300 ms, p70: 1400, p90: 1600 ms, p99: 2100 ms), latency samples: 1720
Latency breakdown for phase 0: ["MempoolToBlockCreation: max: 1.819, avg: 1.725", "ConsensusProposalToOrdered: max: 0.313, avg: 0.307", "ConsensusOrderedToCommit: max: 0.346, avg: 0.331", "ConsensusProposalToCommit: max: 0.654, avg: 0.638"]
Max non-epoch-change gap was: 0 rounds at version 0 (avg 0.00) [limit 4], 1.25s no progress at version 21466 (avg 0.21s) [limit 15].
Max epoch-change gap was: 0 rounds at version 0 (avg 0.00) [limit 4], 0.96s no progress at version 2110574 (avg 0.96s) [limit 16].
Test Ok

• Grafana dashboard
• Humio Logs
• Axiom Logs
• Validator CPU Profile
• Fullnode CPU Profile
• Test runner output
• Test run is land-blocking