sequencer: Extract staging behavior from core

[bobvawter]

This change allows the source of staging data provided to the Core sequencer to
be configured. The StagingCursor type is renamed to BatchCursor and a new
BatchReader interface defines a way to receive an asynchronous stream of cursors.

All existing code is updated to inject a Staging sequencer into the stack. This
sequencer provides an acceptor which writes incoming data to the staging tables
and injects the existing staging-reader implementation into Core.

The Core sequencer's acceptor is now a no-op. The workload checker is updated
to support providing data via BatchReader. This allows Core to be tested
without use of any staging tables.

---

This change is 

[BramGruneir]

Reviewed 31 of 31 files at r1, all commit messages.
Reviewable status: all files reviewed, 4 unresolved discussions (waiting on @bobvawter, @Jeremyyang920, @sravotto, and @ZhouXing19)

---

internal/sequencer/core/core.go line 41 at r1 (raw file):

// The Core sequencer applies transactional data read from a
// [types.BatchReader] in a possibly-concurrent manner.

Can you expland on this comment about possible-concurrency?

---

internal/sequencer/core/core.go line 114 at r1 (raw file):

						_, _, _ = progress.Update(func(stat sequencer.Stat) (sequencer.Stat, error) {
							stat = stat.Copy()
							if len(group.Tables) == 0 {

Is there a test for this?

---

internal/sequencer/staging/acceptor.go line 58 at r1 (raw file):

	ctx context.Context, batch *types.TableBatch, opts *types.AcceptOptions,
) error {
	stager, err := a.stagers.Get(ctx, batch.Table)

Isn't this the same as the inner loop of Accept MultiBatch (with a change in varialbe for stage?
Why not have them both call the same function?

---

internal/sequencer/staging/staging.go line 68 at r1 (raw file):

	opts.Delegate = s.markers.MultiAcceptor(opts.Delegate)

	if err != nil {

Why is this after setting the delegate?

[bobvawter]

Reviewable status: all files reviewed, 4 unresolved discussions (waiting on @BramGruneir, @Jeremyyang920, @sravotto, and @ZhouXing19)

---

internal/sequencer/core/core.go line 41 at r1 (raw file):

Previously, BramGruneir (Bram Gruneir) wrote…

Can you expland on this comment about possible-concurrency?

Done.

---

internal/sequencer/core/core.go line 114 at r1 (raw file):

Previously, BramGruneir (Bram Gruneir) wrote…

Is there a test for this?

The zero-table Group is exercised in the next commit, which switches pglogical over to to using Core.

---

internal/sequencer/staging/acceptor.go line 58 at r1 (raw file):

Previously, BramGruneir (Bram Gruneir) wrote…

Isn't this the same as the inner loop of Accept MultiBatch (with a change in varialbe for stage?
Why not have them both call the same function?

It could be in go 1.23 with the iter package. For the moment, there are enough differences that trying to use a common function was more trouble that it's worth.

---

internal/sequencer/staging/staging.go line 68 at r1 (raw file):

Previously, BramGruneir (Bram Gruneir) wrote…

Why is this after setting the delegate?

Moved the call to Query() to a local variable declaration above the call to Copy().