glTex(Sub)ImageX functions cannot upload float data from memory beyond 2GB address space

Hi,
when using the "-s MAXIMUM_MEMORY=4GB" flag i have run into issues with the glTex(Sub)ImageX functions. As soon as an address beyond the 2GB address space is passed as the pixels parameter and the type parameter is GL_FLOAT the following error is reported: "WebGL: INVALID_OPERATION: texImage2D: ArrayBufferView not big enough for request".
The reason for this issue is the usage of the arithmetic right shift operator (>>) in the generated .js file. When passing an address into e.g. glTexImage2D it is being shifted by 2 bit by the javascript code if float data is supposed to be uploaded. Note that the address is stored as a signed 32 bit signed integer on the javascript side, so any address beyound 2GB will be a negative value. Right shifting negative values via >> will add 1 to the left and thus mess up the intention of the shift.
A potential fix to this problem is modifying the library_webgl.js and library_webgl2.js files in the emscripten repository by replacing several ">>" operators with ">>>" operators. These will be able to handle negative integer values as intended. So for example the line

`GLctx.texImage2D(target, level, internalFormat, width, height, border, format, type, heap, pixels >> heapAccessShiftForWebGLHeap(heap));`

would have to be changed to

`GLctx.texImage2D(target, level, internalFormat, width, height, border, format, type, heap, pixels >>> heapAccessShiftForWebGLHeap(heap));`

There are probably quite a few other places where this change would be necessary to properly support the "-s MAXIMUM_MEMORY=4GB" option. Basically all the texture data upload functions which accept data as something else than a simple byte array will be affected.

I have attached a simple demo program which illustrates this problem
[demo.zip](https://github.com/emscripten-core/emscripten/files/7466574/demo.zip)
It has been built by:
```
emcc (Emscripten gcc/clang-like replacement + linker emulating GNU ld) 2.0.33 (0e68b98e85a1644a7224c360fa6cb21bea5dc33d)
clang version 14.0.0 (https://github.com/llvm/llvm-project 1c05c52de2177a328b7d2d07b697af67eb9f8122)
Target: wasm32-unknown-emscripten
Thread model: posix
```

To build it yourself use something like
`emcc -s MAX_WEBGL_VERSION=2 -s MIN_WEBGL_VERSION=2 -s FULL_ES3=1 -s ALLOW_MEMORY_GROWTH=1 -s MAXIMUM_MEMORY=4GB --memoryprofiler -o test.html test.cpp`

The program initializes opengl via egl, allocates 3GB of memory and then tries to upload float data into a texture which fails with the aforementioned error. If the float array used for upload is being allocated before allocating the 3GB of memory everything works as expected.