misc: use-after-move vlog fixes & remove NOLINTNEXTLINE

[WillemKauf]

Picked out of #30202. The changes to vlog() there must have started picking up on some use-after-moves that aren't currently picked up by the linter because of NOLINTNEXTLINE.

It's a little curious why NOLINTNEXTLINE is there anyways- lets remove it.

Backports Required

• none - not a bug fix
• none - this is a backport
• none - issue does not exist in previous branches
• none - papercut/not impactful enough to backport
• v26.1.x
• v25.3.x
• v25.2.x

Release Notes

• none

[Copilot]

Pull request overview

This PR adjusts several vlog() callsites to avoid bugprone-use-after-move warnings (by snapshotting values before moving the source object) and removes NOLINTNEXTLINE suppressions from the vlog macro definitions.

Changes:

• Snapshot/memoize values (e.g., offsets, sizes, group_id) before moving the underlying objects used to compute them.
• Update log formatting arguments to use the snapshotted values after moves.
• Remove NOLINTNEXTLINE comments from src/v/base/vlog.h macro definitions.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
src/v/transform/rpc/client.cc	Cache ids.size() before ids is moved into an async RPC lambda; use cached size in logs.
src/v/storage/mvlog/segment_appender.cc	Cache batch offsets before moving batch data into the entry body; use cached offsets in logs.
src/v/kafka/server/group_manager.cc	Cache group_id before moving kafka_r into offset_commit; use cached id in subsequent logs.
src/v/cluster/rm_stm.cc	Cache b.base_offset() before moving b into read_fence_batch; use cached offset in log.
src/v/cluster/log_eviction_stm.cc	Log using rp_start_offset/kafka_start_offset variables instead of moved-from val.
src/v/base/vlog.h	Remove NOLINTNEXTLINE suppressions from vlog macro definitions.

Comments suppressed due to low confidence (1)

src/v/kafka/server/group_manager.cc:1081

• first_error is initialized to error_code::none but is only assigned inside if (first_error != error_code::none), so it will never capture the first encountered partition error. This makes the later if (first_error != error_code::none) check ineffective and can cause restore to report success despite per-partition errors. Flip the condition (or otherwise set first_error on the first error) so errors are surfaced correctly.

                      group_id,
                      offsets_ntp,
                      kafka_p.error_code);
                    if (first_error != error_code::none) {
                        first_error = kafka_p.error_code;
                    }

[vbotbuildovich]

CI test results

test results on build#83360

test_status	test_class	test_method	test_arguments	test_kind	job_url	passed	reason	test_history
FLAKY(PASS)	KgoVerifierMultiNodeSelfTest	test_kgo_verifier_multi_node_autoassign	null	integration	https://buildkite.com/redpanda/redpanda/builds/83360#019da192-f0e6-4a3c-82c4-29bfa745c5e2	10/11	Test PASSES after retries.No significant increase in flaky rate(baseline=0.0000, p0=1.0000, reject_threshold=0.0100. adj_baseline=0.1000, p1=0.3487, trust_threshold=0.5000)	https://redpanda.metabaseapp.com/dashboard/87-tests?tab=142-dt-individual-test-history&test_class=KgoVerifierMultiNodeSelfTest&test_method=test_kgo_verifier_multi_node_autoassign

[dotnwat]

wow nice!