Fix Prooph projection manager reuse after reconnect

[lifinsky]

Why is this change proposed?

LazyProophEventStore rebuilds its inner Event Store when the PDO connection changes, but LazyProophProjectionManager keeps returning the manager cached for the same context. That manager still holds the previous Event Store and PDO connection.

After a reconnect, a synchronous projection can therefore read through the old connection while the command appends an event inside a transaction on the new one. The projection cannot see the uncommitted event and leaves the read model unchanged.

Description of Changes

Associate each cached manager with the inner Event Store used to create it and rebuild the manager when that instance changes. This reuses the existing reconnect detection and preserves the per-context cache.

The regression scenario initializes a synchronous projection, reconnects, updates an aggregate inside an open transaction, and checks the read model before commit. It also checks rollback and a subsequent retry.

In-memory mode returns its configured projection manager directly, without reconnect tracking.

Fixes #703.

Pull Request Contribution Terms

• I have read and agree to the contribution terms outlined in CONTRIBUTING.

[lifinsky]

@dgafka, could you please prioritize reviewing and merging this fix for a 1.x patch release, rather than waiting for 2.0?

Applications still running 1.x are affected: after a reconnect, synchronous projections can miss events written inside the current transaction. A 1.x release would let affected applications receive the fix without requiring a major-version upgrade.

Thank you.