Cycles X: Improve performance of transparent shadows with OptiX

This changes the shadow record-all any-hit program to accept all
hits (return without calling `optixIgnoreIntersection`) beyond
the furthest distance recorded after the maximum number of hits
that can be recorded was reached.
OptiX will not call the any-hit program anymore for hits beyond
the distance of the accepted hits and also reduces the current
ray length behind the scenes. As a result performance improves
drastically in scenes where shadow rays can hit a lot of
transparent objects, like the "koro" benchmark scene.

With this applied I now get similar performance with both CUDA
and OptiX in "koro". Not quite perfect yet, but much better than
before.

Reviewed By: brecht

Differential Revision: https://developer.blender.org/D12524

Cycles X: Shading performance improvements by changing inlining behavior for SVM

The shading kernels (shade_surface, ...) are limited by memory a lot. I found several hotspots
where execution was stalled waiting for spills to be loaded back into registers. That's
something that can be adjusted by changing the inlining logic:

For example, the compiler did not inline "kernel_write_denoising_features" (even though it
was marked __inline__), which caused it to force synchronization before the function call.
Forcing it inline avoided that and got rid of that hotspot.

Then there was cubic texture filtering and NanoVDB, which introduced huge code chunks
into each texture sampling evaluation (increasing register and instruction cache pressure),
even though they are rarely actually used. Making them __noinline__ outsources that
overhead to only occur when actually used.

Another case is the SVM. The compiler currently converts the node type switch statement
into a binary searched branch sequence. This means depending on the SVM node hit, the
GPU has to branch over large portions of code, which increases instruction cache pressure
immensely (GPU is fetching lots of code even for stuff it immediately jumps away from
again, while jumping through the binary searched branches). This can be reduced somewhat
by making all the node functions __noinline__, so that the GPU only has to branch over a
bunch of call instructions, rather than all the inlined code.
The SVM "offset" value is passed by value into the node functions now and returned through
function return value, to make the compiler keep it in a register. Otherwise when passed as
a pointer, in OptiX the compiler was forced to move it into local memory (since functions
are compiled separately there, so the compiler is unaware of how that pointer is used).

Differential Revision: https://developer.blender.org/D11816

Enable built-in OptiX curves by default

Starting with OptiX 7.3 curves now behave as expected with Cycles and render the same as the
custom intersection implementation, so enable by default.

Fix T88096: Baking with OptiX and displacement fails

Using displacement runs the shader eval kernel, but since OptiX modules are not loaded when
baking is active, those were not available and therefore failed to launch. This fixes that by falling
back to the CUDA kernels.

Fix missing Cycles CPU name for Arm processors

Cycles: Fix build with OptiX 7.3 SDK

Avoid costly OptiX module creation when only doing denoising

Enable megakernel path in OptiX

Fix shadow any-hit program not building and incorrect index being used

Remove OptiX code that is no longed in use