ContinuousQueries.update() cannot send the fields the server now requires #4

ContinuousQueries.update() and its async twin build the PUT body from only the kwargs passed, and the signature has no database, no source_measurement and no destination_measurement parameter at all:

• src/arc_client/management/continuous_query.py:96-125
• src/arc_client/management/async_continuous_query.py:96-106

PUT /api/v1/continuous_queries/:id replaces the whole definition, and Arc now requires a complete body:

• database became required in Validate continuous query database names arc#995, so update() already returns 400 {"error":"database is required"} against current main — before anything below.
• name, source_measurement, destination_measurement, query (with both placeholders) and interval became required in fix(cq): validate continuous-query measurement names, at the boundary and before each run (#1011) arc#1022. Both land in 26.09.3, so from a released-version standpoint this is one break against 26.09.2.

Anything the body omits was previously written to the row as a blank, which is why the server closed it: a partial PUT silently wiped the stored definition.

What the fix needs

Three new parameters, and a read-modify-write rather than a sparse body: get(query_id), overlay the caller's kwargs, send the full definition. Worth noting the fields that are not required but are still overwritten — description, tag_columns, retention_days, delete_source_after_days, is_active — so a sparse body zeroes them; omitting is_active sets it False and stops the continuous query. Read-modify-write handles all of that in one go.

The documented example shows the already-broken partial call and needs the same change: docs.basekick.net/content/docs/arc/sdks/python/data-management.md:243-246 (and the enterprise copy at :237-241).

arcli is unaffected — it already does read-modify-write.