fix: segfault during cached-block catch-up sync

[ch4r10t33r]

Summary

Fixes two distinct segfaults observed during catch-up sync on x86_64:

1. ssz.serialize corrupting live cached SignedBlocks. Stack bottoms out in
   onBlock iterating aggregated_attestations (or later in verifySignatures)
   with a dangling slice into a cached block's heap buffer. Root cause:
   ssz.serialize on a live cached SignedBlock (directly, or via sszClone)
   mutates nested List / Bitlist fields (aggregation_bits, proof_data),
   so the next read of the same cached block crashes.

2. fetched_blocks hashmap rehash race (folded in from fix race condition leading to seg faults #741). The map in the
   network layer is shared between the libxev main thread (onInterval) and the
   libp2p worker thread (onGossip / onReqRespResponse / onReqRespRequest).
   When the worker inserts and triggers a rehash, the main thread's in-flight
   lookups/iterations become invalid and panic.

Changes

ssz-corruption fix (cherry-picked from fix/libp2p-glue-c-abi-x86_64)

• 1aecb39 chain, database: persist pre-forkchoice signed block SSZ for rocksdb
• 1fd70ea node, types: capture block SSZ bytes at cache time, never re-serialize cached blocks

Highlights:

• types/utils: new sszCloneAndGetBytes — one serialize pass returning both a
  deep clone and the SSZ bytes (caller owns), avoiding a second serialize on the
  same value.
• network: new fetched_block_ssz map (root → []u8) with lifecycle tied to
  cacheFetchedBlock / removeFetchedBlock / deinit; getFetchedBlockSsz /
  storeFetchedBlockSsz helpers.
• node.cacheBlockAndFetchParent: uses sszCloneAndGetBytes and stores the
  bytes alongside the cached block.
• chain.CachedProcessedBlockInfo: new sszBytes: ?[]const u8 field.
• node.processCachedDescendants: retrieves stored bytes and passes them via
  blockInfo.sszBytes.
• chain.onBlock / updateBlockDb: uses blockInfo.sszBytes when present,
  falls back to a disposable-clone serialize only for locally produced blocks.

Hashmap rehash race fix (cherry-picked from #741)

• 71280e3 fix race condition leading to seg faults (by @anshalshukla)

Adds a std.Thread.Mutex to BeamNode that serializes work between the
libxev main thread and the libp2p worker thread. Locks are taken around:

• onGossip
• onReqRespResponse
• onReqRespRequest (.blocks_by_root branch)
• onInterval (chain tick, pending replay, RPC sweep, processReadyCachedBlocks)

Test plan

• zig fmt --check .
• cargo fmt --manifest-path rust/Cargo.toml --all -- --check
• cargo clippy --manifest-path rust/Cargo.toml --workspace -- -D warnings
• zig build test --summary all
• zig build simtest --summary all
• Soak under catch-up-sync scenario on x86_64 hardware to confirm both
  original crashes are resolved

[noopur23]

the pr resolved the seg fault issue.