Update issue templates

Mistral kv quant calibration fixes (#2)

`examples/llama/hf_llama_convert.py -i
cache/baseten_v0.6.1_20231206/repos/mistralai/Mistral-7B-v0.1/1a2e76/dst
-o /tmp/tmp9frqc6v6/c5740d/dst --calibrate-kv-cache -t fp16` fails with
similar errors to smooth quant errors because numpy doesn't support
bfloat16.

Mistral smooth quant fixes (#1)

* hf_llama_convert.py has many places where torch tensor is converted to
numpy and numpy doesn't support bfloat16. Explicit conversion to get
around that.
* Fix issue with loading quantized weights where the calculation looks
off.

Update TensorRT-LLM (NVIDIA#546)

update aarch64 libraries (NVIDIA#525)

revise the homepage (NVIDIA#14)

Co-authored-by: Shi Xiaowei <xiaoweis@nvidia.com>