Consider rephrasing format mapping metadata usage

[lexaknyazev]

Section 5.3 says:

When Appendix B, Mapping of vkFormat values does not have an entry for the value of vkFormat, which will happen for newly added Vulkan formats, the KTX writer must provide any known mapping via the following key-value pairs.

I think there are two issues with this language:

• This text makes the mapping metadata presence normatively depend on the contents of a non-normative Appendix B. Moreover, the Appendix could be modified at any time thus potentially invalidating the existing files.
• This text does not cover the case when the format is representable with DFD and, e.g., OpenGL enums but does not have a Vulkan counterpart. For example, BC5 with reversed order of channels aka ATI2n.

[MarkCallow]

I agree it need tightening. Please propose a PR with the rephrasing.