feat: catch-up for read-only ProviderFactorys

[klkvr]

Implements catch-up logic for read-only ProviderFactory.

It is implemented like the following:

reth/crates/storage/provider/src/providers/database/mod.rs

Lines 269 to 297 in 6ea9ce8

	/// For read-only factories, checks whether the MDBX committed txn ID has advanced since the
	/// last sync and, if so, catches up the RocksDB secondary instance and re-initializes the
	/// static file index.
	///
	/// The fast path (txnid unchanged) is lock-free. Only the actual sync I/O is serialized.
	///
	/// No-op for read-write factories.
	pub fn sync_providers_if_needed(&self) -> ProviderResult<()> {
	let Some(sync_state) = &self.read_only_sync else { return Ok(()) };
	let current_txnid = self.db.last_txnid().unwrap_or(0);
	// Fast path: no contention when nothing changed.
	if current_txnid == sync_state.last_synced_txnid.load(Ordering::Relaxed) {
	return Ok(());
	}
	// Slow path: serialize the actual catch-up I/O.
	let _guard = sync_state.sync_lock.lock().unwrap_or_else(|e| e.into_inner());
	// Double-check after acquiring the lock — another thread may have already synced.
	if current_txnid == sync_state.last_synced_txnid.load(Ordering::Relaxed) {
	return Ok(());
	}
	self.rocksdb_provider.try_catch_up_with_primary()?;
	self.static_file_provider.initialize_index()?;
	sync_state.last_synced_txnid.store(current_txnid, Ordering::Relaxed);
	Ok(())
	}

This function is invoked in 2 cases:

1. For every created DatabaseProvider or before every read directly from static files (those bypass mdbx entirely)
2. Every time when mdbx dir changes — this part is identical to static files watcher we have right now and is kept to make sure that read-only factories don't fall too far behind when unused, causing potentially long catch up to happen on next access

[github-actions]

⚠️ Changelog not found.

A changelog entry is required before merging. We've generated a suggested changelog based on your changes:

Preview

---
reth-provider: minor
reth-chain-state: minor
reth-db: minor
reth-engine-tree: patch
reth-cli-commands: patch
reth-e2e-test-utils: patch
---

Added on-demand MDBX syncing for read-only `ProviderFactory` instances via a new `with_read_only_sync` method and `ReadOnlySyncState` tracker. Refactored `ReadOnlyConfig` to watch the MDBX directory instead of static files (renamed `watch_static_files` to `watch`, simplified `StaticFileProvider::read_only` signature). Enhanced `TestBlockBuilder` with a `with_state` mode that generates proper `BundleState` with account/storage changes and reverts, and added a comprehensive reorg consistency test for read-only providers.

Add changelog to commit this to your branch.

[joshieDo]

we might be able to use this in tempo test suite as well cc @SuperFluffy

[joshieDo]

should we be more specific that this should be used for out of process concurrent access?

[joshieDo]

should we be more specific that this should be used for out of process concurrent access?