Em-nesting, alignment-nesting, greek arithmetic numbering

[Omikhleia]

Proposed features from silex.sile - precisely, the subset from silex.fixes

Motivated by #1866

• Nesting-aware \em (Does SILE (intend to) support nesting \em? #1048) - breaking
• Nesting-aware (i.e. honoring the margins) center, raggedright and raggedleft + the latter two no longer cancelling paragraph indent¹ - breaking
  • The "blockquote" example in the manual is fixed to also properly honor alignment, and the command is provided in the plain class
• Greek alphabetical numbering special case (similar to "alpha").
• Small robustness fix/refactor for inputter options

Each topic is a separate commit, with detailed commit logs/

Still to do if we agree on the solution:

• Check failing tests
• Add new tests
• Check the manual (obvious change done in the above commits, but maybe we can improve other places)

Footnotes

1. This is a bold move, but the two things are somewhat orthogonal here, and parindent-cancellation is doable by other means. ↩

[Omikhleia]

This is a bold move, but the two things are somewhat orthogonal here, and parindent-cancellation is doable by other means

But i'm not strongly attached to it either. It makes things simpler when one would nest various alignments, but that's kind of an edge case anyway.

[alerque]

A cursory glance suggests I'll be up for most of this. I targeted it for v0.15.0 since I'd like to resolve especially the inputter changes by then as well as bathing as many breaking changes as we can without making migration too hard. If we cherry-pick some and not other bits that's fine too, but lets try to sort out yes now/no never/maybe later answers on breaking changes sooner rather than later.

[alerque]

I cherry picked the obvious wins of the inputter touchup and Greek numbering (with a couple new tests) to develop directly, that should leave just the intertwined margin/nesting stuff for this PR.

[alerque]

I think we'll move ahead with the current breaking change that does not cancel parindent, but we'll add a shim to the retrograde package that restores the old behavior. This should be relatively easy to shim this way, and since the actual value of parindent is also changing in this release I think it makes sense to allow the breakage but also support old documents with the same mechanism.

Obviously the retrograde shim will not be expected to also support nested use cases the way the new code does, just cover the very basic case of a single level usage and simulate the old behavior. This will probably mean either stashing a private setting, having some private hook, or registering a wrapper/replacement function.