Consider two I-Frame only playlists, to allow adaption based on frame rate the two playlist have different I-Frame duration.

The I-Frame only segment has a single sample (IDR), of sub-second duration (e.g. 0.0333s for 30fps source) but a
duration equal to the time distance between the I-Frames.

Playlist P1 -- 6 seconds between IDR samples, and Playlist P2 -- 4 seconds between IDR samples:

````
+-----+-----+-----+
|  1  |  2  |  3  |
+-----+-----+-----+
+---+---+---+---+---+
| 1 | 2 | 3 | 4 | 5 |
+---+---+---+---+---+
````

If the `AdaptiveTrackSelection` changes from P2 after loading segments from P1 there appears to be a sample overlap that
would trigger `shouldSpliceIn` to be set on the iFrame `MediaChunk` when no sample splicing is required.

````
+-----+---
|  1  |  2    --- loaded in P1
+-----+---
        +---+---+---+
        | 3 | 4 | 5 |  -- switched to P2
        +---+---+---+
````

Because the `startTimeUs` of P2 segment 3 is less then the `endTimeUs` of P1 segment 3 it appears samples would overlap,
as there is only one sample in each segment this is not possible.   Setting this `shouldSpliceIn` flag incorrectly
prevents pruning the buffered chunks.

Add test cases to cover the exiting logic for `shouldSpliceIn(...)` including two cases that cover:

1. Current pull request proposed changes (Intra-iframe only track switch
2. Proposed logic for to/from iFrame only track