Chao Zhang

~~Looks like there might be issues with the `:` and `...` arguments as well, when used on the first dim of the tensor.~~ ~~Caused by #769~~

This will block resolving #768

Ah yes, I stand corrected. It looks like 0.4.0 should still work as previously described; the issue should be in master only. Let me give your suggestions above a try...

WIP but somewhat working advanced indexing. Needs to be based off of https://github.com/NVIDIA-AI-IOT/torch2trt/pull/770

@jaybdub This is ready for review, but I may need to rebase... I started this before 0.4.0 was released, and certainly before the major API changes to master.

Note that this is dependent on #770

Also note that I've left in a couple of TODOs... I'll leave those for future PRs

```NUM_TESTS: 88 NUM_SUCCESSFUL_CONVERSION: 88 NUM_FAILED_CONVERSION: 0 NUM_ABOVE_TOLERANCE: 0 NUM_pSNR_TOLERANCE: 0 ```

From my local testing looks like this implementation should now correctly support dynamic shapes as well.

@jaybdub This PR is ready for review