storage: remove unnecessary sideload inlining, add assertion

[nvanbenschoten]

Related to #17500.

Replica.sendRaftMessages used to contain logic to inline sideloaded entries
for MsgApps. This was unnecessary, as MsgApps would never contain thin entries.
This is because replicaRaftStorage.Entries already performs the sideload inlining
for stable entries and raft.unstable always contain fat entries. Since these are
the only two sources that raft.sendAppend gathers entries from to populate MsgApps,
we should never see thin entries in MsgApps.

This change replaces the inlining attempt with an assertion.

Release note: None

[cockroach-teamcity]

This change is 

[bdarnell]

---

Review status: 0 of 2 files reviewed at latest revision, 1 unresolved discussion, all commit checks successful.

---

pkg/storage/replica.go, line 4180 at r1 (raw file):

		case raftpb.MsgApp:
			// FOR REVIEW: this assertion will be quick for all non-sideloaded
			// commands. Is it worth always performing the assertion?

I could go either way but I guess I'd lean towards keeping this assertion RaceEnabled-only.

---

Comments from Reviewable

[tbg]

and thanks!

---

Reviewed 2 of 2 files at r1.
Review status: all files reviewed at latest revision, 1 unresolved discussion, all commit checks successful.

---

pkg/storage/replica.go, line 4180 at r1 (raw file):

Previously, bdarnell (Ben Darnell) wrote…

I could go either way but I guess I'd lean towards keeping this assertion RaceEnabled-only.

What Ben said. The recipient would find out pretty quickly that they're screwed, too.

---

Comments from Reviewable

[nvanbenschoten]

Review status: 1 of 2 files reviewed at latest revision, 1 unresolved discussion.

---

pkg/storage/replica.go, line 4180 at r1 (raw file):

Previously, tschottdorf (Tobias Schottdorf) wrote…

What Ben said. The recipient would find out pretty quickly that they're screwed, too.

Done.

---

Comments from Reviewable