Consumes any-sdk v0.5.3-alpha10, which makes ColumnDescriptor.GetName()
return the display name consistently across the resource-schema and
response/select-items-schema paths (wire name only via GetWireName).

Fixes the postgres backend regression where materialized table/view
columns were named by the wire name while SELECT projected the display
name, so a case-sensitive backend rejected them (42703) - covering the
xml:name, snake_case_aliases and request.nativeCasing triggers.

Verified on a real postgres backend: #108 multi-word values populate,
SELECT * no longer errors, #109 snake WHERE resolves, #105 OData
push-down intact. No stackql source change required.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

any-sdk keys the row payload by the schema property key. That key equals
GetWireName when the wire name carries the case difference (snake_case_aliases:
display cidr_block over wire cidrBlock) but equals GetName when the display name
does (an xml:name override: display AvailabilityZone over wire availabilityZone).

The #108 change extracted solely by GetWireName, which fixed the snake_case case
but regressed the xml:name case (e.g. aws.ec2.volumes AvailabilityZone) to NULL on
alpha10 - visible only on a case-sensitive backend once alpha10 fixed the column
name. Try GetWireName first, then fall back to GetName, so both resolve their
value. Undecorated columns have GetWireName == GetName, so they are unchanged.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Reverts next commit. Emits DRMDBG to stderr (captured by the robot exec keyword ->
output.xml) with the column GetName/GetWireName and the actual payload keys when a
value cannot be extracted, so we can see what key the plain-XML objectKey walker used
for aws.ec2.volumes AvailabilityZone (misses both GetName and GetWireName).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The prior debug logged on every value-extraction miss, including genuinely-null
JSON columns, which flooded stderr and broke the Go integration tests (go test) in
the Build step - blocking the downstream Test jobs (robot) we need for the aws line.

Now it only logs when the column is decorated (GetName != GetWireName, i.e.
snake_case_aliases or an xml:name override) AND still misses. Undecorated columns
(all JSON providers) never log, so go test / Build pass and the robot Test jobs run,
emitting DRMDBG for aws.ec2.volumes AvailabilityZone. Reverted with the fix.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Adds actions/upload-artifact steps after the mocked functional robot runs (winbuild
native + docker postgres_tcp) so output.xml / log.html / report.html survive the
runner. This lets us inspect the exec stderr (e.g. the temporary DRMDBG payload-key
dump) after a run instead of losing it when the runner is torn down. if: always() so
it uploads on failure too. Temporary aid for the aws.ec2.volumes value diagnosis.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Root cause of the CI "Server Parameter in Select Star" failure was a stale test
expectation, not a stackql defect. The aws.ec2.volumes AvailabilityZone column
carries an allOf-nested `xml: name: availabilityZone` alias; the drm value
extraction now keys the response payload by GetWireName (1d97800), so the column
correctly projects `ap-southeast-1a`. The test at line 2362 still asserted the old
pre-fix `null`. The snake-aliased sibling views (volumes_presented / _naively_ /
_poorly_) already expect `ap-southeast-1a`, corroborating the correct value.

- stackql_mocked_from_cmd_line.robot: AvailabilityZone null -> ap-southeast-1a (both rows).
- drm_cfg.go: revert the temporary payload-key debug (6cdafd8, 5734373) back to the
  clean GetWireName-then-GetName extraction.
- build.yml: revert the temporary robot-report upload-artifact steps.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>