Clarify UNPACK_PREMULTIPLY_ALPHA for floating-point and integer pixel formats

The `UNPACK_PREMULTIPLY_ALPHA_WEBGL` parameter is defined in WebGL 1.0 spec that on its own supports only unsigned normalized pixel formats.

The WebGL 2.0 and the related WebGL extension specs should define how this parameter works with:
- `RGBA` and `LUMINANCE_ALPHA` formats with `FLOAT` and `HALF_FLOAT` types. Specifically, how floating-point overflow/underflow is handled.
- `RGBA_INTEGER` format. Specifically, how signed/unsigned integer overflow is handled.
- `RGBA` format with `BYTE` and `SHORT` (from `EXT_texture_norm16`) types.

No files or tests are identified in the issue. Start by reviewing the WebGL 2.0 and related WebGL extension specifications, then trace the definition of UNPACK_PREMULTIPLY_ALPHA_WEBGL for each listed format and type. Done means the specifications explicitly define overflow and underflow behavior for all requested cases.

Written by the indexing model from the issue text.