FileStore with segment_count 1 loses every record written after its first compaction #17

Version: master at d5282e6; also 0.1.7, and master before #6.
Platform: all (host reproduction, RAM filesystem from test_file_store).

What happens. Compaction writes the live records into seg0, and both rotate_segment_if_needed() and compact_if_threshold() then set current_segment = 1 for new writes. With segment_count == 1, seg1 is outside the configured range. Every later write goes there, and the file grows without bound, because rotation now compacts only seg0, which is empty. rebuild_index_from_segments() and compaction scan only [0, segment_count), so after a reload those records are gone.

How to reproduce: init(fs, "/p", false, 256, 1), then overwrite ten keys six times (60 puts of ~20 bytes), then reopen. Expected: 10 records with their last values. Actual: 1 record; seg0 is 0 bytes and seg1 is 2420 bytes. The same workload with segment_count == 2 is correct.

Impact: low. The default is 8 segments and microReticulum's documented minimum is 3, but nothing rejects 1, and the failure is silent data loss.