[SYCL] Correct spec about constness of get(properties_tag) member and add respective warning

[HPS-1]

Currently some of our docs suggest using get(properties_tag) {...} member in kernel functors to specify kernel properties while some suggest using get(properties_tag) const {...}. But actually get(properties_tag) {...} might not work under certain conditions, so we should update the incorrect docs that they suggest readers to use get(properties_tag) const {...} rather than get(properties_tag) {...}.

Also as @Pennycook suggested, added a diagnostic warning to indicate the user when:
-they have declared a get(properties_tag) member in the kernel functor
AND
-they declared get(properties_tag){...} rather than get(properties_tag) const {...}

Updated the tests containing the non-const version of get(properties_tag) as well.

[kbenzie]

UR LGTM

[Bensuo]

Graph changes LGTM!

[HPS-1]

Requesting @intel/llvm-gatekeepers to merge this. Thanks!

[aelovikov-intel]

@Pennycook , I think there was some other PR where it was decided that kernel properties can't be runtime. Should we change this to static constexpr instead?

[aelovikov-intel]

Adding @gmlueck and @rolandschulz per @Pennycook 's suggestion.

Also note that we need to modify

llvm/sycl/include/sycl/handler.hpp

Lines 339 to 340 in daca6a3

	template <typename TransformedArgType, int Dims, typename KernelType>
	class RoundedRangeKernel {

to propagate the properties, making this constexpr static would make that much easier.

[gmlueck]

I agree that kernel properties will always be compile-time. For a case where a runtime property is required, I think we would use a launch property instead. Therefore, it seems like it's OK to make this static constexpr.

[Pennycook]

I found the discussion @aelovikov-intel was referencing over in #17633, where we ended up adding both static constexpr and auto versions to deal with compile-time and run-time properties.

Committing to only compile-time properties would solve the problem here and simplify that extension, so I'm also happy to make this static constexpr.