Correct Javadoc warnings, prevent future ones

[niloc132]

Addresses several classes of issues, some in more than one way:

• Ensure all javadoc runs use UTF-8 encoding, and remove non-ASCII chars which are intended to be interpreted as code
• HTML-encode characters that must be encoded in javadoc, and correct wrongly-encoded chars.
• Don't reference JVM internals that can't be linked, instead refer to them as @code.
• Fixed mismatched html tags, broken attrs.
• Corrected/removed classes/members that don't exist or can't be referenced in that way.

Fixes #1513

[niloc132]

This is left as a draft for the moment to get some feedback from a few specific authors/owners. Only combined-javadoc will fail if there are warnings, allowing potentially "invalid" javadoc references that can be resolved by producing javadoc all at once.

[rcaudy]

Why 11?

[devinrsmith]

I'm guessing b/c that's the language level we target.

[niloc132]

Yes, just centralized the javadoc configuration to one place, so everything is built the same way, rather than 3-ish different setups.

[rcaudy]

Dropping -quiet was intentional?

[niloc132]

Yes, it did nothing, it was a workaround for a gradle javadoc limitation. Boolean is probably more accurate or at least less confusing, so no one thinks that -quiet was intended to be added as a real flag.

[devinrsmith]

buildSrc changes look good. I did go fast through the actual javadoc changes. I call this a big win, even if we feel like there are places we might want to follow-up on to make further improvements. I would not hold up this PR for javadoc content-type changes.

[devinrsmith]

This was the only file that stood out as having some sort of new code - I'm assuming you re-ran the generation?

[chipkent]

Same question.

[niloc132]

Codegen was updated, else the build would fail when the assert check ran.

[chipkent]

Same question.