Add Purge Now action to Manage Servers (closes #900)

[erikdarlingdata]

Summary

• Adds a Purge Now button to the Manage Servers window with a confirm dialog and mode picker (Use configured / 1 / 3 / 7 / Custom / All).
• Right-click menu on the Manage Servers grid now mirrors every per-row action (Edit, Toggle Favorite, Check Server Version, Purge Now, Remove) above the existing Copy/Export items.
• Reworks config.data_retention semantics so @retention_days does what users expect from the UI:
  • NULL (default) — respect per-collector retention from config.collection_schedule, 30-day fallback for unscheduled tables.
  • 0 — TRUNCATE every collect.* table.
  • N > 0 — override every table's cutoff to N days, ignoring per-collector schedule.
• The old @truncate_all parameter was removed; @retention_days = 0 covers that case with cleaner semantics.
• Truncate path snapshots SUM(partition.rows) before wiping so the result message can show "Rows wiped: 914,983" instead of "Rows wiped: 0".

Test plan

• EXEC config.data_retention (NULL) on sql2019 → uses schedule, deletes 0 rows when nothing is older than 30 days.
• EXEC config.data_retention @retention_days = 5 on sql2019 → deleted 184,481 rows (override skipped the schedule pass).
• EXEC config.data_retention @retention_days = 0 on sql2019 → 914,983 rows wiped, 52 tables truncated, 31ms; log row records both counts.
• FK / schema-bound / indexed-view check on sql2016 — none reference collect.*, TRUNCATE is safe.
• Updated proc applied and smoke-tested on sql2016, sql2017, sql2019, sql2022, sql2025.
• Dashboard build clean; UI tested manually for all dialog modes including the second confirm on All.
• Manual test on a fresh installer run to verify no proc-deployment regression.

Closes #900.