Grid Column Filters: Distinct "Null" Option, Count-Column Filtering Fix, and Rules Lookup Reliability

A batch promotion to QA that improves how grid column filter dropdowns behave. Filters can now offer a distinct "Null" (empty) option separate from 0, count columns such as document and mail counts now populate their filter values correctly, and a rules-engine caching defect that could break lookups was fixed.

Several grid filter dropdowns were not working as expected. NULL/empty values were silently dropped, so users could not filter for records that simply had no value (and an empty cell looked the same as a 0). Count columns (e.g. document count, mail count) failed to list any filter values because the configured column key (document_count) did not match the key actually present in the data payload (total_documents). Separately, the rules service was calling the cache manager with arguments in the wrong order, which could cause folder/label/priority/status/tag lookups to misbehave.

• Users can now explicitly filter grid columns for empty/missing values via a distinct "Null" choice, clearly separated from numeric 0.
• Filter dropdowns for count columns (document count, mail count) now show real values instead of appearing empty.
• Rule configuration screens that resolve folders, labels, priorities, statuses, and tags are more reliable, and rule action values can carry both legacy text and enriched id/name details.
• No user-facing UI redesign; these are correctness improvements to existing filtering and rules behavior.

Backend (FastAPI) changes, promoted as a Dev -> QA batch ("Batch Document Count"):

• column_values_service.py: stop filtering out NULLs when building filter values; track explicit nulls and append None to the result so the UI can render a distinct "Null" option; sort with null first for both numeric and non-numeric types; adjust has_more to account for the null entry.
• Added a _synonym_keys helper mapping equivalent column keys: document_count<->total_documents, mail_count<->total_mails, status<->status_name; item lookups now try the requested key then its synonyms.
• response_flattener.py: extract_value_by_column_key now falls back to the same synonym keys when config and payload key names differ.
• rules_service.py: fixed cache_manager.get/set call signatures (key/namespace order, expire= -> ttl=) across folder, label, priority, status, and tag lookups.
• rule_schemas.py: widened ActionValueResponse.value to accept legacy str/list[str] plus enriched id/name dict forms.
• _format_values updated to preserve None instead of discarding it; doc comment in grid_columns.py updated to reflect null inclusion.

Low to moderate. The largest behavioral change is that filter value lists now include NULL, which alters dropdown contents and sort order (null-first); the frontend must render None as a "Null" label rather than blank. The synonym key fallback is additive and only triggers when the primary key is absent, but could surface values for previously-empty count filters. The rules cache fix corrects a real argument-order bug; the main risk is exercising previously dead/broken cache paths. Mitigation: QA verification on QA before any production promotion.

• Open grid filter dropdowns and confirm a distinct "Null" option appears for columns that have empty values, and that it is separate from 0.
• Verify count-column filters (document count, mail count) now list values instead of being empty; confirm sorting (null first, then ascending/descending) is correct.
• Confirm search-within-filter still works and respects the row limit/has_more flag including the null entry.
• Exercise rule configuration: verify folder, label, priority, status, and tag lookups load correctly and are cached (no errors on repeat loads).
• Confirm rule action values render for both simple string/list values and enriched id/name objects.