Add memory limit for buffered repl changes

[morgo]

A Pull Request should be associated with an Issue.

We wish to have discussions in Issues. A single issue may be targeted by multiple PRs.
If you're offering a new feature or fixing anything, we'd like to know beforehand in Issues,
and potentially we'll be able to point development in a particular direction.
Further notes in https://github.com/block/spirit/blob/main/.github/CONTRIBUTING.md

Fixes #822

[Copilot]

Pull request overview

Adds a per-subscription “soft” memory cap to buffered replication subscriptions to prevent wide-row workloads from growing the in-memory buffer without bound, and updates client locking to avoid deadlocks when backpressure causes HasChanged to block.

Changes:

• Track approximate buffered bytes per subscription and apply backpressure by parking HasChanged on a condition variable once the soft limit is reached.
• Make the per-subscription soft limit configurable via ClientConfig.SubscriptionSoftLimitBytes (default 256 MiB; negative disables).
• Adjust client lock usage around subscription dispatch (and DDL matching) to avoid deadlocks when HasChanged parks.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
pkg/repl/subscription_buffered.go	Adds byte accounting, soft-limit backpressure via sync.Cond, and logs/counters for parking.
pkg/repl/subscription_buffered_test.go	Initializes cond in existing tests and adds new unit/integration tests for accounting + park/wake behavior and deadlock regression.
pkg/repl/README.md	Documents the new memory backpressure behavior and operational implications.
pkg/repl/client.go	Introduces default/configured soft limit, wires cond during subscription construction, and releases c.Lock before calling HasChanged.
pkg/repl/client_test.go	Updates tests to initialize cond for manually constructed subscriptions.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.