fix: Automatic detection and backfilling of gaps in block history at startup

[mouradski]

Problem:

During batch indexing, if a block fails to insert (timeout, constraint violation, etc.), it is silently ignored due to INSERT IGNORE.

On application restart, the indexer would simply resume after MAX(number) from the blocks table, leaving permanent gaps in the history.

Example:
Blocks in DB: [1000, 1001, 1002, 1005, 1006] ← blocks 1003-1004 missing
Previous behavior: restart → starts at 1007
Result: blocks 1003-1004 are permanently lost

Solution:

• Added a continuity check at startup in getStartIndex():
• Retrieve statistics: MIN(number), MAX(number) and COUNT(*)
• Gap detection: compare expected vs actual block count
• Locate first gap: SQL query with NOT EXISTS to find the first missing block
• Smart resume: indexer resumes from the first gap instead of max

Robustness:
✅ Executed once at startup → negligible cost
✅ Self-healing → after a crash, the next restart automatically fills gaps
✅ Performance → SQL query is O(n log n) with index on number
✅ Backward compatible → unchanged behavior if no gaps detected