Refactor all shaders to handle Vulkan implementations with 64K max thread blocks in X axis

[azonenberg]

This breaks things on Intel.

[azonenberg]

Current operating ranges of shaders along each grid axis. If checked, we should be good on any reasonable GPU (64K minimum thread block size and a screen resolution << 64K)

(Counts displayed below are total thread count required in each axis, not thread block count)

• WaveformToneMap: x=window width, y = window height
• ConstellationToneMap: x=window width, y = window height
• EyeToneMap: x=window width, y = window height
• WaterfallToneMap: x=window width, y = window height
• SpectrogramToneMap: x=window width, y = window height
• ScopeDeskewUniform4xRate: x=search range (default 30K)
• ScopeDeskewUniformEqualRate: x=search range (default 30K)
• ScopeDeskewUniformUnequalRate: x=search range (default 30K)
• waveform-compute: x = window width, y = fixed 64 threads

[azonenberg]

Looks like we actually did all of them, closing.