activation: add metal implemenation

[drbh]

This PR adds a metal implementation of various activation functions that are comparable with the cuda version.

This should helps make the current basic usage example work on osx https://huggingface.co/docs/kernels/en/basic-usage and in general enable osx users to use activation kernels

[drbh]

update to include the bench output

tested with the changes to the kernels benchmark command in this pr huggingface/kernels#265

kernels benchmark path:/Users/drbh/Projects/kernels-community/activation
Downloading path:/Users/drbh/Projects/kernels-community/activation@local...
Running benchmark.py...
Kernels library path: /Users/drbh/Projects/kernels/kernels/src/kernels/__init__.py

  GPU      Apple M3 Max (30 cores)
  CPU      Apple M3 Max
  OS       Darwin 25.2.0
  PyTorch  2.10.0

  Running SiluWorkloads on mps

┌───────────────┬────────────┬─────┬───────────┬────────────┬───────────┬───────────┬───────────┬───────────┬────────────┬───────────┬─────────┐
│ Benchmark     │ Workload   │   N │ Speedup   │   Mean(ms) │   Std(ms) │   Min(ms) │   Max(ms) │   IQR(ms) │   Outliers │   Ref(ms) │ Match   │
├───────────────┼────────────┼─────┼───────────┼────────────┼───────────┼───────────┼───────────┼───────────┼────────────┼───────────┼─────────┤
│ SiluWorkloads │ large      │ 100 │ 1.72x     │     6.6684 │    0.6442 │    6.2739 │   10.6387 │    0.2859 │         11 │   11.4963 │ ✓       │
│ SiluWorkloads │ medium     │ 100 │ 2.67x     │     1.213  │    0.4494 │    1.027  │    4.0239 │    0.0895 │          7 │    3.2332 │ ✓       │
│ SiluWorkloads │ small      │ 100 │ 1.87x     │     0.5019 │    0.2439 │    0.4133 │    2.8096 │    0.0258 │         11 │    0.9388 │ ✓       │
└───────────────┴────────────┴─────┴───────────┴────────────┴───────────┴───────────┴───────────┴───────────┴────────────┴───────────┴─────────┘

  large: 1.72x faster (95% CI: 6.5421-6.7947ms vs ref 11.4963ms) ✓ significant
  medium: 2.67x faster (95% CI: 1.1249-1.3011ms vs ref 3.2332ms) ✓ significant
  small: 1.87x faster (95% CI: 0.4541-0.5497ms vs ref 0.9388ms) ✓ significant

Kernel: 5f9348f  Benchmark: 760b882

[MekkCyber]

lgtm! thank you