Fix Data Race in Epoch Boundary

[nisdas]

What type of PR is this?

Bug Fix

What does this PR do? Why is it needed?

WARNING: DATA RACE
Write at 0x00c0b095b278 by goroutine 70382:
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).updateEpochBoundaryCaches()
      beacon-chain/blockchain/process_block.go:329 +0x509
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).handleEpochBoundary()
      beacon-chain/blockchain/process_block.go:360 +0x29a
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).lateBlockTasks()
      beacon-chain/blockchain/process_block.go:611 +0x74a
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).runLateBlockTasks()
      beacon-chain/blockchain/process_block.go:480 +0x3a4
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).Start.func2()
      beacon-chain/blockchain/service.go:214 +0x33

Previous read at 0x00c0b095b278 by goroutine 571785:
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).updateEpochBoundaryCaches.func1()
      beacon-chain/blockchain/process_block.go:316 +0xb9

Goroutine 70382 (running) created at:
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).Start()
      beacon-chain/blockchain/service.go:214 +0x304
  github.com/prysmaticlabs/prysm/v5/runtime.(*ServiceRegistry).StartAll.func1()
      runtime/service_registry.go:46 +0x42

Goroutine 571785 (running) created at:
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).updateEpochBoundaryCaches()
      beacon-chain/blockchain/process_block.go:310 +0x257
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).handleEpochBoundary()
      beacon-chain/blockchain/process_block.go:360 +0x29a
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).lateBlockTasks()
      beacon-chain/blockchain/process_block.go:611 +0x74a
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).runLateBlockTasks()
      beacon-chain/blockchain/process_block.go:480 +0x3a4
  github.com/prysmaticlabs/prysm/v5/beacon-chain/blockchain.(*Service).Start.func2()
      beacon-chain/blockchain/service.go:214 +0x33

Which issues(s) does this PR fix?

There have been reports of odd deadlocks or high resource consumption on prysm nodes. In the above reported race, it is possible for the main goroutine to execute before the child goroutine, leading to the epoch being 1 lesser than the one we want. If this happens we will miss caching the next epoch's shuffling.

Other notes for review

[potuz]

LGTM