Added IsNan to shaders w/out boilerplate

[TheJeran]

The 2D shaders dont use a boilerplate which is where I was storing the isNan bitcheck function. So those shaders weren't working

[gemini-code-assist]

Code Review

This pull request injects the ${isNaNFunc} helper into several WGSL shader templates to facilitate NaN handling. While this allows the shaders to compile, the feedback highlights significant logic errors in the reduction, covariance, and correlation shaders: although NaNs are skipped during summation, the final calculations still divide by the total dimension length rather than a count of valid samples, leading to incorrect statistical results. Additionally, the helper is injected into the CUMSUM3D shader without being used, and the isNaN implementation's behavior of treating Infinity as NaN should be verified against the intended requirements.

[gemini-code-assist]

The addition of ${isNaNFunc} allows this shader to compile, but the logic inside the main function (lines 555-556) is incorrect when NaNs are present. The shader currently skips NaNs in the summation loop but still divides by the total dimLength to calculate the mean. This will result in an incorrect average. You should implement a counter to track the number of valid (non-NaN) samples and use that for the division.

Also, note that the isNaN implementation (line 15) returns true for both NaN and Infinity. While often desirable in data processing, ensure this matches your requirements as the function name suggests it only checks for NaN.

[gemini-code-assist]

Similar to the other reduction shaders, the logic in main (lines 667-668 and 682) is incorrect when NaNs are skipped. It uses dimLength for averaging and covariance calculation, which includes the skipped indices. A dynamic counter for valid samples should be used instead to ensure statistical correctness.

[gemini-code-assist]

The logic in this shader (line 774) uses dimLength as the sample count N. Since NaNs are skipped during the summation loops, N should be a count of the actual valid pairs processed to avoid incorrect correlation results.

[gemini-code-assist]

The isNaN function is defined here but is not used anywhere in the CUMSUM3D shader body. If you don't intend to filter NaNs during the accumulation, this line should be removed to keep the shader code clean.

[TheJeran]

Will address the concerns later