Feature/9348 improve download queue

[shaunahu]

Tag each generating-data-file-for-* download job with a fair-share class, small or large, taken from the existing size estimate. A download estimated under 50MB (aws.batch.job.share.small-max-size) is small. Anything else is large: 50MB or more, a failed estimate, or the size limit switched off.

• The tag is always sent as the share_identifier job parameter. DAS can read it, and parquet downloads carry it into their prepare/collect child jobs.
• It is also sent as the top-level Batch shareIdentifier, but only when aws.batch.job.share.enabled=true. That defaults to false, so no environment changes behaviour: AWS rejects shareIdentifier on a queue without a fair-share policy.

Follow-up, once infra has a queue with a fair-share scheduling policy whose share identifiers are small and large:

• ogcapi: set share.enabled=true and point aws.batch.job.queue / child-queue at that queue.
• DAS: set shareIdentifier on the prepare/collect child jobs it submits.

[utas-raymondng]

LGTM