Add support for visualizing self-attention heatmaps + sequence classifier outputs w/ attentions (#1117)
* Add support for nearest neighbour interpolation
* Support bigint inputs for `min` and `max` helper functions
* Add min/max/argmin/argmax tensor ops
* Allow sequence classifier outputs with attentions
* Align depth estimation pipeline w/ python version
* Update `min` and `max` types
* Fix qwen2vl processor
* Update depth estimation pipeline unit tests
* Remove old comments
* Update types
* Fix type issues