To get some misc asciidoc fixes.

Load data before trigger priority was ambiguous.

Make the data of the *context registers 32 bits wide.

Fix contradictory req. for ndmresetpending

Not just M-Mode and D-Mode.

Clarify bit recommendations on scontext and mcontext.

Clarify textra behavior when XLEN changes.

Remove Section 4.9: XLEN

Section 4.1 already says "Effective XLEN is DXLEN." The statement about
communicating XLEN to the user is unnecessary to the spec, and unclear,
so we don't need it.

Fix asciidoc conversion for DXLEN definition.

Fix minor AsciiDoc conversion problem.

Dual-license XML files BSD-2-Clause and CC-BY-4.0

Ensure a newline at the end of registers description

Without this, the `[cols=...]` line would show up in the rendered text
(at the start of section 4.10).

Conflicts:
	debug_module.tex

The rule makes those targets from scratch now.

Fix `hartset` -> {hartsel}

Fixes #972

Also multiple identifiers must be on the same line together.

BSD-2-Clause is used so the C code generated from those files can be
used in GPLv2 open source projects (specifically, OpenOCD) and the
CC-BY-4.0 clause is the standard license for work produced by the RISC-V
Foundation.

Convert to asciidoc

Once this merges, I'll make a github release including the PDF, so the
first link will work.

Fix text overlapping in offsets in register diagrams.

bytefield really handles this very poorly, so this is a hack by moving
long offsets in a bit. Now some look a little odd with not much space
between upper and lower offset, but at least there's no actual
overlapping text anymore.

Top of some characters would sometimes be cut off otherwise.

E.g. if the field is "hit" and the offset is "XLEN-18" then the field
needs to be sized to accommodate that offset.

Remove stray $ left over from LaTeX.

Instead explicitly list both fields.

Fixes #964.

Misc asciidoc fixes.

It never parses right. I can't figure out how to do this properly.