[RFC] Lowering `aten.index_add_` through the TM Tensor Dialect

Nobody has claimed this yet.

Hey @ramiro050 and @cathyzhyi,

Trying to move `aten.index_add_` so that it goes through the TM Tensor Dialect

Inspired by [this PR lowering aten.bincount](https://github.com/llvm/torch-mlir/pull/574/commits/f818ce22f3999e28b05455b188aaa24dc9ee1798), here's my current plan:

1. Add `AtenIndexAdd` to `e2e_testing/torchscript/basic.py`
2. Define `Toch_AtenIndexAddOp` in `include/torch-mlir/Dialect/Torch/IR/GeneratedAtenOps.td`
3. Create a `AtenIndexAddOp` class in `lib/Conversion/TorchToTMTensor/TorchToTMTensor.cpp`
-- I'm also going to the op under `runOnOperation` within the `ConvertTorchToTmTensor` class
4. Add the op to the conditional statement in `lib/Dialect/Torch/Transforms/RefineTypes.cpp`
5. Emit the op under `python/torch_mlir/dialects/torch/importer/jit_ir/build_tools/torch_ods_gen.py`

Would love any feedback! Am I thinking about the sequence correctly?

Thank you in advance,

Lorenzo

Start with e2e_testing/torchscript/basic.py and the referenced aten.bincount lowering, then inspect the listed Torch dialect, TorchToTMTensor, RefineTypes.cpp, and torch_ods_gen.py locations. Done means aten.index_add_ is represented and lowered through the TM Tensor Dialect, with the end-to-end test entry added.

Written by the indexing model from the issue text.