feat(ingest): --continuous flag on new-datasource (buckets)

[eddietejeda]

Summary

Adds a --continuous flag to hotdata ingest new-datasource (buckets/filesystem only). It rides through to the datasource request body's continuous field, marking the datasource for the native cross-workspace continuous-sync scheduler — the dlthubworker-sync CronJob that hits /internal/ingest/jobs/sync-all every ~60s and incrementally re-drains every continuous filesystem datasource (append-only on newly-arrived object keys).

Until now the flag existed on the worker/API contract but there was no CLI way to set it — you had to hand-craft a raw POST /ingest/sources. This closes that gap.

Why

We land external telemetry into S3 buckets and read it back through Hotdata (e.g. the MQTT and the new OTel otel.hotdata.dev → s3://hotdata-labs-otel-telemetry-…/otel pipelines). Those buckets get new objects continuously, so their datasources should auto-refresh rather than needing a manual trigger-import each time. --continuous is the one-liner that opts a bucket datasource into that.

hotdata ingest new-datasource --service buckets \
  --bucket-url s3://hotdata-labs-otel-telemetry-857481120914/otel \
  --format jsonl --glob '**/*.json' --name otel_telemetry \
  --config @creds.json --continuous

Changes

• src/commands/ingest.rs — --continuous arg on new-datasource, wired into the created request; unit test create_request_filesystem_carries_continuous_flag asserts it rides through (and is off by default).
• src/client/ingest.rs — continuous: bool on IngestRequest (skipped when false, so the wire contract is unchanged for everyone else).
• skills/hotdata/SKILL.md — document the flag.

Notes

• Filesystem-only. The worker only honours continuous for family=filesystem; it's ignored for SQL/REST/Iceberg/etc. (matching the worker's _validate_ingest).
• Backward compatible — the field is omitted from the body unless --continuous is passed.