Non-max-merging for Inference Slicer

[LinasKo]

Description

This PR builds on top of @mario-dg PR #500, replacing it, fixing issues, adding non-max-merging to Detections and connecting to InferenceSlicer.

There main stylistic complexity here is dealing with individual objects inside Detections.

Here, the following changes are implemented:

• Detections now have with_nmm, which merges the detections based on their overlap.
• detections/core.py exposes an OverlapFiltering enum, which allows other methods to distinguish between the types. This enum is exposed globally in __init__.py.
• InferenceSlicer is the only class using it right now. NMS is performed by default, and users may specify overlap_filtering to pick which filtering type to use.
• New method Detections._set_at_index mimicks the former __setitem__, to be used internally. Only used in one location, so happy to refactor there.
• Removed Optional from InferenceSlicer's iou_threshold arg.

⚠️ So far, only tested with object detections - not segmentation.
⚠️ Naming could be made more clear, e.g. in methods like utils.py::batch_non_max_merge. Ping me and I'll look into that.

Type of change

Please delete options that are not relevant.

• New feature (non-breaking change which adds functionality)
• This change requires a documentation update

How has this change been tested, please provide a testcase or example of how you tested the change?

Tests:

• Works with 0 detections
• Works with ~5 detections
• Works with inference slicer, visual results the same as drawn in SAHI example in Add Non-Maximum Merging (NMM) to Detections #500
• Benchmarking in progress

Colab: https://colab.research.google.com/drive/1iiJ5oCl8dwwcefRgNXRdaFH6G9dVpf4P?usp=sharing

Any specific deployment considerations

One thing I couldn't figure out was the right way to access values of individual detected objects.
When a scalar was needed, I went with self.xyxy[i]. When an array was expected, I left the self[i].xyxy calls. Feels a bit backwards, but most concise - is there a better way?

Docs

• Docs updated? What were the changes:

[LinasKo]

Colab: https://colab.research.google.com/drive/1iiJ5oCl8dwwcefRgNXRdaFH6G9dVpf4P?usp=sharing

[LinasKo]

It fails when Detections.merge is called on segmentation detections. Frequently mask=None is returned, so detections in the Slicer can't be merged.

My guess is that it's caused by my old friend #943, but I have not verified.

[LinasKo]

Closing; Moving fixes to #500 as discussed