fix: reset fragmentation state on seek to prevent invalid fragment ranges

[lideen]

Description

After upgrading this library from 0.5.6 to the current version 2.3.0, I was intermittently seeing errors where ISOFile.createFragment() was called with an invalid sample range (sampleEnd < sampleStart), especially around seek operations.

Root cause: seek() updated trak.nextSample, but per-track fragmentation/extraction state was not reset.
So after a seek, state could be out of sync (for example, stale lastFragmentSampleNumber from before seek vs new nextSample after seek).

This could cause:

• Invalid fragment ranges (sampleEnd < sampleStart) after backward seeks
• Fragment windows incorrectly spanning pre-seek and post-seek samples after forward seeks
• Unnecessary/incorrect fetch offsets (requesting earlier data than needed)
• Segment buffer contamination across seek boundaries

Solution

On each track seek (seekTrack), reset seek-sensitive per-track state:

• set lastFragmentSampleNumber to nextSample
• set lastSegmentSampleNumber to nextSample
• reset accumulatedSize to 0
• clear segmentStream
• clear pending extracted samples for extraction mode

Also added a defensive guard in createFragment:

• if sampleEnd < sampleStart, skip fragment creation safely and warn