Simplify Library Rescan

[imnotjames]

Problem Statement

The library rescan logic is duplicated between multiple locations - in particular with a manual rescan vs turning on the watch feature and dropping a file. This leads to two different behaviors and two locations where we'd need to benchmark and optimize to handle larger libraries more effectively.

Goals

• Simplify Library Rescan: Use the same logic between a manual library rescan and library watching
• Handle "modified" files: On watch or during rescan, we should detect when a known file has changed
• Smaller Transactions / Lower Memory Usage: Attempt to compartmentalize this logic into smaller commit units to improve performance on larger libraries
• Simplify Regression Testing: Have only one spot where we need to validate regressions

Non-Goals

• Handle failures relating to too many notify listeners: This is a known problem and something we can't really control
• Handle watch logic across network: While we could in theory make the rescans happen regularly for watch logic across the network this seems kinda outside the scope

Proposed Solution

Move the library rescan code & the monitoring services over to the watch services.

Then, on rescan emit watcher events to the watcher processEvents:

• If file is not a currently known book path, emit an ENTRY_CREATE
• If file should exist but is now deleted, emit an ENTRY_DELETE
• If the file checksum has changed, emit an ENTRY_MODIFY

Then, merge in all of the rescan logic we can into the file event processor.

Ensure transaction boundaries on each event rather than the entire rescan.