Update on "Arm backend: Add 16A8W support and test for tanh operation"

Add 16A8W quantization support and test for the tanh operation in ExecutorTorch ARM backend.

This follows the pattern established for linear, mul, and sigmoid operations, extending int16 support to tanh operations.

Changes:
- Add INT16 dtype validation support in op_tanh.py
- Add test_tanh_tensor_16a8w_tosa_INT test function
- Enable test_tanh.py in test targets configuration

The 16A8W configuration uses 16-bit activations with 8-bit weights, enabling higher precision for activations while maintaining weight efficiency.

Differential Revision: [D80510815](https://our.internmc.facebook.com/intern/diff/D80510815/)

cc digantdesai freddan80 per zingo oscarandersson8218

[ghstack-poisoned]

Author: Ninja91