fix: add txlookup lock to prevent reorg race and deadlock (ethereum#29343, #34039)

[hmlee-wm]

Summary

Ports ethereum#29343 and ethereum#34039. During reorg, a concurrent GetTransactionLookup() call could read stale DB data between Purge() and the new chain write, caching a wrong block hash/number. Additionally, #29343 introduced txLookupLock without defer, leaving early-error return paths in reorg() unlocked and causing permanent deadlock.

Upstream References

PR	Description
#29343	core: add txlookup lock
#34039	core: fix txLookupLock mutex leak on error returns in reorg()

Problem

Race condition (#29343): Purge() was called before writing new chain data. A concurrent RPC call hitting GetTransactionLookup() during this window read stale reorged data from DB and cached it, causing eth_getTransactionReceipt / eth_getTransaction to return wrong block hash and number until node restart.

Deadlock (#34039): reorg() acquired txLookupLock.Lock() but released it only on the happy path. Early returns on errInvalidOldChain / errInvalidNewChain leaked the lock, permanently deadlocking all future transaction lookups and reorg attempts.

Changes

• core/blockchain.go: add txLookupLock sync.RWMutex to BlockChain; in reorg(), acquire lock before mutation and use defer Unlock() to protect all return paths; move Purge() to after DB writes complete
• core/blockchain_reader.go: add txLookupLock.RLock() / defer RUnlock() in GetTransactionLookup() to block reads during reorg

[colinkim]

LGTM

[eomti-wm]

LGTM