linkchecker:analyze gets stuck when analyzing nodes deleted via trash module

Problem/Motivation

When using the Trash module,
deleted entities are stamped with a timestamp in the deleted column
of their data table. The getQuery() method in LinkExtractorBatch
does not filter out these entities, so they are returned as candidates for extraction.
However, loadMultiple() does not return soft-deleted entities, causing a
mismatch between the IDs found by the query and the entities actually loaded.

This results in an infinite loop during drush linkchecker:analyze or
the UI batch process, where the same unloadable entity IDs are returned on every batch
run but never processed, preventing the batch from ever finishing.

The same infinite loop can also occur with other types of database inconsistencies
where an entity exists in the index tables but cannot be loaded by
loadMultiple().

Steps to reproduce

1. Install the Trash module.
2. Delete one or more nodes (move them to trash).
3. Run drush linkchecker:analyze.
4. The process gets stuck and never reaches 100%.

Proposed resolution

Fixes in LinkExtractorBatch:

1. Filter deleted entities in getQuery(): Check
   if the entity's data table has a deleted column (indicating the Trash
   module or similar is installed) and add an IS NULL condition to exclude
   deleted entities from extraction candidates.
2. Handle unloadable entities: If loadMultiple()
   returns fewer entities than IDs were found, mark the missing IDs as indexed in
   linkchecker_index and log a warning. This prevents an infinite loop
   for any type of DB inconsistency and gives administrators visibility into the problem.
3. Add infinite loop guard to batchProcessEntities():
   If processEntities() returns 0 processed items, mark the batch as
   finished regardless of the total count. This is a safety net to prevent infinite
   loops from any future edge case.

Remaining tasks

• Review and testing