schedule: Add more restriction to descript gpu

[yadej]

Motivation

Some gpu schedule should not be possible so I add more validation at the loop nest level
Among the checking

• We need gpu_block to take an axis
• We need gpu_thread, gpu_lane or gpu_warp to be a tile
• We can't have a loop before the gpu_block because that does not pass the compilation part but most importantly it would be making a operator making a lot of kernel

[codecov]

Codecov Report

❌ Patch coverage is 19.23077% with 21 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/xtc/schedules/parameter_loop_nest.py	7.14%	13 Missing ⚠️
src/xtc/schedules/loop_nest.py	33.33%	4 Missing and 4 partials ⚠️

📢 Thoughts on this report? Let us know!