Use timeout to allow RpcClient to retry initial transaction confirmation

[CriesofCarrots]

Problem

In a load-balanced RPC setup with imperfect or nonexistent session affinity, an RpcClient may query a recent-blockhash and send a transaction to one node, and attempt to confirm the transaction on another. If this second node is behind or on a short-lived fork, it may not have yet seen the blockhash, causing the client to exit early here:

solana/client/src/rpc_client.rs

Lines 1858 to 1864 in 9d983a3

	if self
	.get_fee_calculator_for_blockhash_with_commitment(
	recent_blockhash,
	CommitmentConfig::processed(),
	)?
	.value
	.is_none()

even though the transaction may have actually landed.

Summary of Changes

Add a configurable timeout for initial transaction confirmation, allowing get_signature_status retries even when the blockhash cannot be found. In a single-node RPC setup, this shouldn't add any additional time.
Also, a little cleanup in RpcClient to use the commitment() method (this is most of the churn), and refactoring in RpcClient to make this cleaner to use and make adding additional RpcClient config parameters easier in the future.

[t-nelson]

LGTM! Cleanups look great! Comments below should be taken as nits 😉

Pull request has been modified.

[codecov]

Codecov Report

Merging #18311 (7e747f0) into master (9d983a3) will increase coverage by 0.0%.
The diff coverage is 55.7%.

@@           Coverage Diff           @@
##           master   #18311   +/-   ##
=======================================
  Coverage    82.3%    82.3%           
=======================================
  Files         433      433           
  Lines      121160   121193   +33     
=======================================
+ Hits        99793    99846   +53     
+ Misses      21367    21347   -20