Migrate to more precise terminology for "old-world and new-world" topics

[xen0n]

While being convenient to use, the terms "old world" / "new world" are oftentimes too vague, and this can lead to people conflating different aspects of the problem. The Loongson terms "ABI1.0" / "ABI2.0" are no better choice because the semantics is equivalent to the former, and arguably the mention of "ABI" is worse, because "object file ABI" is already a thing when it comes to ABI differences.

So, we should adopt a new approach where we clearly distinguish the different aspects of LoongArch ABI differences:

Aspect	New term in favor of "new world"	New term in favor of "old world"
Boot protocol	UEFI	BPI
Linux UAPI	(the) 5.19 UAPI	(the) "4.19" UAPI1
Object file ABI	preferably unqualified, "object file ABI v1", or EF_LARCH_OBJABI_V1 if necessary	"object file ABI v0"2
Relocation style	preferably unqualified, "direct-style relocs" if necessary	stack-machine-based relocs

We should add info boxes to remind people to be aware of this, and switch to the new, more precise terminology for newly written material.

Footnotes

1. Quote the "4.19" part whenever possible, because the upstream 4.19.x versions of Linux actually does NOT support LoongArch. ↩

2. There is no EF_LARCH_OBJABI_V0. ↩

[xen0n]

Give Copilot a try, this is not going to be final ;-)

[xry111]

IMHO "object file ABI" and "relocation style" should be considered the same concept. "EF_LARCH_OBJABI_V1 is not set" should mean "the object file ABI / relocation style is unknown," while EF_LARCH_OBJABI_V1 should mean "the object file ABI is v1 and no stack-based relocation types are used."