Atomic support for enhancement

[metesynnada]

Which issue does this PR close?

Closes 101 and address multiple TODOs inside the codebase.

Rationale for this change

If a large number of states undergo updates, using RwLock for the scheduler's in-memory states at the global level will reduce the overall performance. Additionally, some state updates are not atomic and were listed as TODOs. The modifications in this PR makes concurrency more efficient during heavily loaded in-memory updates and more reliable during state persistence.

What changes are included in this PR?

• Changed RwLock<HashMap<K, V>> into a DashMap. The dashmap crate provides an implementation of a more sophisticated, high-performance concurrent hash map. For reference, DashMap stats are

  Stars	Forks	Used By	Contributors	License
  1.8k stars	104 forks	42.8k	34	MIT license

• Now performing atomic transactions on TaskManager and ExecuterManager, resolving TODOs.

  • Change in trait StateBackendClient, converted method put_txn to a more general apply_txn.
  • Rollback if the state persisting is not successful.

Are there any user-facing changes?

No

[thinkharderdev]

LGTM

[thinkharderdev]

nit

Suggested change

	pub async fn with_locks<Out, F: Future<Output = Out>>(
	locks: Vec<Box<dyn Lock>>,
	op: F,
	) -> Out {
	let mut locks = locks;
	let result = op.await;
	locks.reverse();
	pub async fn with_locks<Out, F: Future<Output = Out>>(
	mut locks: Vec<Box<dyn Lock>>,
	op: F,
	) -> Out {
	let result = op.await;
	locks.reverse();

[ozankabak]

Thanks for the review, just sent a commit that incorporates your suggestion.