db,wal: avoid Batch.refData allocation

[jbowens]

Passing in the Batch.refData func into (wal.Writer).WriteRecord forced an allocation for every committed batch. This commit defines a new wal.RefCount interface implemented by Batch to avoid the allocation.

Informs cockroachdb/cockroach#123236.

[cockroach-teamcity]

This change is 

[jbowens]

Reviewable status: 0 of 6 files reviewed, 1 unresolved discussion (waiting on @aadityasondhi, @RaduBerinde, and @sumeerbhola)

---

batch.go line 232 at r1 (raw file):

const batchClosedBit = 1 << 30

// TODO(jackson): Hide the wal.RefCount implementation from the public Batch interface.

Any ideas on how to achieve this?

[sumeerbhola]

Reviewed 6 of 6 files at r1, all commit messages.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @aadityasondhi, @jbowens, and @RaduBerinde)

---

batch.go line 232 at r1 (raw file):

Previously, jbowens (Jackson Owens) wrote…

Any ideas on how to achieve this?

Not really. Wrapping Batch in another struct seems unnecessary boilerplate.

---

wal/wal.go line 389 at r1 (raw file):

	// like to reuse p's memory, the caller may pass a non-nil [RefCount].  If
	// the Writer will retain p, it will invoke the [RefCount] before returning.
	// When it's finished, it will invoke the func returned by the [RefCount] to

nit: func returned by the RefCount is stale.

[jbowens]

TFTR!

Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @aadityasondhi, @RaduBerinde, and @sumeerbhola)

---

wal/wal.go line 389 at r1 (raw file):

Previously, sumeerbhola wrote…

nit: func returned by the RefCount is stale.

good catch