…rations

The GEMM overview table in inductor logs was missing batch size information
for batched matrix operations like torch.bmm and torch.baddbmm. This made it
difficult to distinguish between different batched operations with the same
M, N, K dimensions but different batch sizes.

Changes:
- Updated counter key format in kernel files to use prefixed values
  (e.g., "aten.bmm_b10_m1024_n1024_k1024" instead of "aten.bmm_10_1024_1024_1024")
- Enhanced parsing logic in compile_fx.py to handle both new prefixed format
  and legacy format for backward compatibility
- Added batch size display in overview table for batched operations
  (e.g., "aten.bmm (B=10)" instead of just "aten.bmm")
- Increased table width to accommodate batch size information

Before:
```
Name                 | M    | N    | K    | Count
aten.bmm            | 1024 | 1024 | 1024 | 1
```

After:
```
Name                           | M    | N    | K    | Count
aten.bmm (B=10)               | 1024 | 1024 | 1024 | 1
```

This provides clearer visibility into batched GEMM operations while maintaining
backward compatibility with existing counter formats.

Fixes pytorch#155307