Sampler TODOs

[NicolasHug]

Opening this issue keep track of outstanding tasks that were left as follow-up from #221

• Implement different policies for handling edge-case when the clip span is larger than the number of available frames https://github.com/pytorch/torchcodec/blob/6850e189ee5902f505af163dd7f966507a63d7a7/src/torchcodec/samplers/_implem.py#L99-L103
• Implement fast and efficient "sort and decode indices" strategy https://github.com/pytorch/torchcodec/blob/6850e189ee5902f505af163dd7f966507a63d7a7/src/torchcodec/samplers/_implem.py#L120-L124
• Decide whether we should return List[FrameBatch] or FrameBatch
• Move FrameBatch and Frame to top-level (requires small refactoring)
• Rename _implem.py into _implementation.py