158341: distmerge: add merge processor scaffolding r=mw5h a=mw5h

#### bulkmerge: add processors and DistSQL physical planning

This commit introduces the initial implementation of bulkmerge processors
for distributed merge operations. It adds three new processor types:
MergeLoopback, BulkMerge, and MergeCoordinator, along with the DistSQL
physical planning infrastructure to orchestrate them.

The MergeLoopback processor runs on the coordinator node and generates
initial tasks. The BulkMerge processors run on each SQL instance to
perform merge operations. The MergeCoordinator collects results from
all merge processors.

This commit also updates the vectorized execution engine (execplan.go)
to recognize the new processor cores, preventing panics when these
processors are encountered.

Informs #156580
Release note: None

#### backup: deduplicate routing logic with physicalplan package

Previously, the backup package contained duplicate implementations of
routingDatumsForSQLInstance and routingSpanForSQLInstance that were
identical to functions recently added to pkg/sql/physicalplan. This
commit removes these duplicates and uses the shared implementations:
RoutingDatumsForSQLInstance and MakeInstanceRouter.

This reduces code duplication and ensures consistent routing behavior
across DistSQL physical planning for both backup/restore and bulk merge
operations.

Informs #156580
Release note: None

#### physicalplan: set DefaultDest to nil to catch routing bugs

Changes MakeInstanceRouter to set DefaultDest to nil instead of stream 0.
When DefaultDest is nil, any routing key that doesn't match a span will
produce an error rather than silently routing to an arbitrary stream.
This helps catch coordination bugs early.

Updates merge_loopback.go to generate a routing key for its SQL instance
using physicalplan.RoutingDatumsForSQLInstance instead of a hardcoded
"loopback" key. This ensures the routing key matches one of the spans in
the router rather than relying on the DefaultDest fallback.

The test expectation is updated from "loopback->merge->coordinator" to
"node1->merge->coordinator" to reflect the explicit routing behavior.

Fixes #156580
Release note: None


158398: ui, sql: Record KV CPU Time in statement and transaction statistics r=alyshanjahani-crl a=alyshanjahani-crl

This commit starts recording the newly added kvCPUTime field from topLevelQueryStats and exposes it in the UI.

Fixes: https://cockroachlabs.atlassian.net/browse/CRDB-57265 Release note (ui change): KV CPU Time is now recorded to statement_statistics and transaction_statistics and is displayed in the SQL Activity page.

158538: mmaprototype: add more logging to candidate selection decisions r=angeladietz a=angeladietz

This adds more logging to help understand decisions MMA is making around when a lease/replica can be moved to another store.

Informs #158203
Epic: CRDB-55052
Release note: None

158595: ui: fix double dispatch when metric time is selected r=dhartunian a=dhartunian

Previously, when a new time window was selected in the time picker on the metrics page, we would dispatch a `ts/query` request twice because the first request would trigger a `useEffect` callback that triggered a second one. This was because of a desire to keep query params in the URL bar in sync with the time picker. We split up the query param update with the query param + redux update into two functions so that we can sync the params withou touching redux. This eliminates the double request cycle since redux is now updated once.

Resolves: #158507
Epic: None

Release note (bug fix): the DB Console no longer dispatches duplicate metric query requests when new time windows were selected on the metrics page.

158596: kvnemesis: correctly verify the span config of the system range r=miraradeva a=miraradeva

Previously, kvnemesis verified the correct span config for some critical ranges: meta (r1), liveness (r2) and system (r3). However, as of 8844739, meta1 and meta2 start off as separate ranges, pushing the system range to r4. It's important for this range to be replicated correctly in kvnemesis to ensure it's available in the presence of network partitions. Unavailability of the system range can result in splits not able to allocate new range IDs.

This commit bumps up the max range ID used for span config verification from 3 to 4.

Informs: #158366

Release note: None

158606: mma: always incorporate RangeMsg.RangeLoad r=wenyihu6 a=sumeerbhola

Epic: CRDB-55052

Release note: None

158610: mma: tweak the comment of MakeStoreLoadMsg r=wenyihu6 a=sumeerbhola

This is mainly to clarify that we sum the load and capacity reported per store inside MMA, so tweaking the capacity calculation approach may not be possible in isolation.

Epic: CRDB-55052

Release note: None

158611: sql: increase TestScatterWithOneVoter observability r=iskettaneh a=iskettaneh

This commit traces the scatter statement, and adds some logging to help investigate future failures of this test.

Fixes: #158491

Release note: None

Co-authored-by: Jeff Swenson <[email protected]>
Co-authored-by: Matt White <[email protected]>
Co-authored-by: alyshanjahani-crl <[email protected]>
Co-authored-by: Angela Dietz <[email protected]>
Co-authored-by: David Hartunian <[email protected]>
Co-authored-by: Mira Radeva <[email protected]>
Co-authored-by: sumeerbhola <[email protected]>
Co-authored-by: iskettaneh <[email protected]>