A line comment may end the input (#1039)

[Rafael-SOWNet]

Fixes #1039.

COMMENT in AngouriMath.g was

COMMENT: ('//' ~[\r\n]* '\r'? '\n' | '/*' .*? '*/') -> skip ;

so the // alternative only matched when a newline followed. A comment on the last line never
matched at all and its first / reached the parser as an operator. The end of the input is where a
comment is most likely to be written — the last line of a file, or a single-line string handed to
FromString — and the block form carries no such condition, so this reads as a missing optional
rather than a decision.

COMMENT: ('//' ~[\r\n]* ('\r'? '\n')? | '/*' .*? '*/') -> skip ;

Measured on a build of this branch:

input	before	now
x + 1 // done	UnhandledParseException: extraneous input '/'	x + 1
x + 1 //	same	x + 1
x + 1 // a // b	same	x + 1
x + 1 // done\n	x + 1	x + 1
x + 1 // done\r\n	x + 1	x + 1
x + 1 /* done */	x + 1	x + 1
x + // c\n 1	x + 1	x + 1
x / 1	x / 1	x / 1

// nothing on its own is still a parse error — the comment is skipped and an empty input is not an
expression, which is the same reason "" is an error.

Regeneration

Per Docs/Contributing/ImproveParser.md: the vendored antlr-4.13.1-complete.jar, then the
AntlrPostProcessorReplacePublicWithInternal step that makes the generated classes internal. Worth
noting for review — only the lexer changes. Regeneration first rewrites AngouriMathListener.cs,
AngouriMathBaseListener.cs and AngouriMathParser.cs with public in place of internal, and the
post-processor puts them back byte-identical, so the diff is the grammar, AngouriMathLexer.cs and
AngouriMathLexer.interp and nothing else.

One behaviour change beyond the fix

"//" on its own is still an error, and its reported position moves from line 1:0 to line 1:2.
That is the convention the neighbouring FromStringTest rows already follow — " " is line 1:2,
"/**/" is line 1:4, both pointing at the end of the skipped text — so // stops being the one
row that pointed at a character the lexer was meant to have consumed. The row is updated with a
comment saying why.

Tests

SyntaxDocumentedTest.ALineCommentNeedsItsNewline pinned the old behaviour, with #1039 linked
against it. It becomes ACommentMayEndTheInput over six spellings, plus
ACommentIsNotAnExpressionOnItsOwn. Docs/Usage/Syntax.md said the newline was required and now
says it is not, keeping the issue reference for anyone reading an older version.

Suite: 8107 passed, 0 failed, 14 skipped.

Part of #746 item 8 — this was found while deriving what the grammar accepts.