fix(ci): sds-elastic e2e setup tolerates already-consumed block devices

[nevermarine]

Description

The sds-elastic e2e setup (configure-sds-elastic.sh + wait-sds-elastic.sh) waited for >= N BlockDevices with .status.consumable == true before creating the ElasticCluster. On a pipeline rerun the raw OSD disks are already consumed into sds-elastic OSD LVMVolumeGroups (sds-elastic-<ec>-osd-*), so they are no longer consumable and the wait looped until timeout:

Error:  Consumable blockdevices did not reach 6 in time

This PR makes the setup idempotent across restarts:

• A BlockDevice now counts as an OSD candidate when it is either still consumable (fresh run) or already backing an sds-elastic OSD LVG (rerun), detected via .status.lvmVolumeGroupName matching ^sds-elastic-.*-osd- (override with ELASTIC_OSD_LVG_REGEX). Real provisioning state is used rather than the user label, so detection survives a BlockDevice rediscovery.
• The app=elastic-osd labelling step (the ElasticCluster.spec.storage.blockDeviceSelector) uses the same predicate, so it re-labels already-consumed devices too — self-healing if a label was lost during rediscovery.
• Shared predicate (ELASTIC_OSD_BD_SELECT) and regex live in wait-sds-elastic.sh and are reused by both scripts; readiness log/error messages renamed Consumable → OSD-candidate.

Why do we need it, and what problem does it solve?

Reruns of the nightly sds-elastic e2e pipeline (or a rerun after a partial provisioning) always failed at the block-device wait, because consumed disks never return to consumable. This unblocks pipeline restarts.

What is the expected result?

The block-device readiness check passes on:

• fresh run — all disks consumable;
• full rerun — all disks already consumed into OSD LVGs;
• partial rerun — a mix of consumable and already-consumed disks.

Verified the jq predicate on fixtures (consumable + elastic-consumed counted, unrelated LVGs and plain non-consumable disks excluded) and bash -n on both scripts. A live-cluster check was not possible — the referenced e2e cluster had already rotated.

Checklist

• The code is covered by unit tests.
• e2e tests passed.
• Documentation updated according to the changes.
• Changes were tested (jq predicate + shell syntax); not user-facing.

Changelog entries

section: ci
type: fix
summary: sds-elastic e2e setup tolerates already-consumed block devices on pipeline rerun.
impact_level: low