Parity: range-partition a document table by a non-tenant column (e.g. by date) — SQL Server equivalent of marten#4779 #211

Context

Marten #4779 (PR marten#4780) covers declaratively range-partitioning a single-tenant (or any) document table by an arbitrary, non-tenant column — the classic time-series retention pattern, e.g. a duplicated DateTimeOffset (BucketEnd) range-partitioned monthly so pruning old data is an instant DROP/SWITCH of a partition instead of a large DELETE.

On the Marten/PostgreSQL side the public entry point already existed (DocumentMappingExpression<T>.PartitionOn(member, cfg) over a duplicated column); the only real work was a Weasel partition-bound round-trip fix (weasel#318, shipped in Weasel.Postgresql 9.3.0).

Ask

Provide the equivalent for Polecat (SQL Server) so the document API has parity:

• A way to declaratively partition a non-tenant document table on a caller-chosen (duplicated) member, including range partitioning by a date column.
• SQL Server partitioning differs structurally from PostgreSQL (partition functions + schemes + a partitioning column, with SWITCH/SPLIT/MERGE RANGE for rolling partitions) rather than declarative child tables, so the API shape and migration story will differ from Marten's.

Notes

• Worth checking whether Weasel.SqlServer's partitioning types need the same round-trip/normalization hardening that Weasel.Postgresql got in Converge event-store auxiliary operations (archive / tombstone / progression) onto shared Weasel.Storage.Events seams #318 before wiring up the document-side API.
• Externally-managed / scheduler-rolled partitions are likely the most useful first cut here too.
• Filed as the SQL Server companion follow-up explicitly called out in marten#4779.