threads: set stack pointer correctly for x86-64

[lsf37]

The tutorial was setting the stack pointer on x86-64 such that it was not aligned to 16 bytes. gcc 14.2 emits SSE instructions that require 16 byte alignment and fault when the stack is not correctly aligned.

The misalignment stems from the fact that x86-64 expects a return address on the stack when the function is entered, but we were passing the top of the stack instead. The Arm ABI has the return address in the link register instead.

Explain this difference in the tutorial and adjust the code accordingly. The later helper function sel4utils_arch_init_local_context already did this directly.

[midnightveil]

Is that TODO for you or the tutorial user? (the latter seems a bit weird?).

[lsf37]

Is that TODO for you or the tutorial user? (the latter seems a bit weird?).

It's for the tutorial user, the tutorials all use TODO tags to mark the bits in the template that should be filled in.

This one is indeed a bit weird, but it was worse before, because the text, the template, and the solution all said different things about whether or not to use sel4utils_arch_init_local_context function. Introducing the sel4utils_arch_init_local_context function earlier would theoretically work, but then the stack pointer issue would be hidden and would start to hit people if they actually started using things.