Improve VerifyBlockError::Commit typing, so we don't accidentally break syncer error handling

[conradoplg]

Motivation

VerifyBlockError::Commit is a wrapper over a BoxError. However, in #2890 a check was added in the should_restart_sync function to ignore Commit errors corresponding to "block is already committed to the state" when deciding whether to reset the sync procedure. That check is fragile since it will break if the string changes.

Change Commit to wrap a specific error type enumeration, and create a specific item for that particular error to be used for matching and filtering it in should_restart_sync.

Currently, this check is implemented by the is_duplicate_request() method, which can't use Commit because it is a BoxError.

There's also a BlockDownloadVerifyError::Invalid error that comes from the chain verifier and a BlockDownloadVerifyError::DownloadFailed from the network service.

We should downcast all BoxErrors in the syncer's BlockDownloadVerifyError to concrete types and add them to the match statement in should_restart_sync():

zebra/zebrad/src/components/sync.rs

Line 1088 in 0152c86

fn should_restart_sync(e: &BlockDownloadVerifyError) -> bool {

zebra/zebrad/src/components/sync.rs

Lines 1155 to 1163 in 0152c86

	// TODO: add a proper test and remove this
	// https://github.com/ZcashFoundation/zebra/issues/2909
	let err_str = format!("{e:?}");
	if err_str.contains("AlreadyVerified")
	|| err_str.contains("AlreadyInState")
	|| err_str.contains("block is already committed to the state")
	|| err_str.contains("block has already been sent to be committed to the state")
	|| err_str.contains("NotFound")
	{

Specifications

Designs

• See Improves block commit error typing (WIP) #5682
• See fix(lint): Fix result_large_err lint by making errors smaller, improve error handling #5757
  • we don't need to reorder fields manually to save space, because Rust 1.67 will do this automatically: optimize field ordering by grouping m*2^n-sized fields with equivalently aligned ones rust-lang/rust#102750
  • we can still turn unnamed tuple fields into named fields, because that makes debugging easier

Related Work

• Dedupe Verification error types (Block, Chain) #1186

Follow up to #2890

[mpguerra]

Hey team! Please add your planning poker estimate with ZenHub @conradoplg @dconnolly @jvff @oxarbitrage @teor2345 @upbqdn

[teor2345]

We're de-prioritizing network changes, and network-related changes.

[teor2345]

This is needed to do #5487