Decouple and unify PEP header processing for rendering, PEP 0, JSON, RSS and linting

[CAM-Gerlach]

As a followup to #2584 and #2585 and as I've already been thinking about lately and discussed with @JelleZijlstra and @warsaw at PyCon, right now we parse the headers three different places with three different sets of logic:

• pep_zero_generator.parser for PEP 0, the JSON and the RSS with some additional sub-parsing in those callers
• pep_processor.transforms.header_transform for the PEP rendering
• The .pre-commit-config.yaml pygrep hooks for linting),

Instead, we should just use the structured format I suggest in #2584 (with the parsing presumably in the PEP class) for all of them (starting with PEP rendering initially), for several reasons:

• Simpler, more DRY, reliable and maintainable in the long run
• Easier to add new headers in one central place, as in #2572
• Enables simplifying/loosening the allowed input syntax for authors without fear of breaking anything
• Simplifies adding requested features in the various outputs (e.g. :abbr: from #2514 )
• Decouples the header format from reST/Sphinx and ensures the format-independence @AA-Turner originally wanted in #2358
• Avoids cryptic regexes for linting (as @hugovk requested on #2484)

Of course, we should decide on and implement the structured format proposed in #2584 first, as it is a perquisite for this (alongside the changes in #2585 and #2579 needing to be merged before work on this begins), but I've opened this to keep track of it.