parser: fix failures found with fuzzing

[spytheman]

• parser: make the parser a bit more robust for fuzzed source code, and show more context on scanner fails for source code without backing .v files
• parser: make the parser more robust for fuzzed source code

[spytheman]

The used file was x_fuzz.v.txt
produced by editing examples/2048/2048.v so that [inline] was renamed to @[inline], then running:
zzuf -R '\x00-\x20\x7f-\xff' -r0.01 < examples/2048/2048.v > x_fuzz.v
Then v -g -d trace_parse_stmt cmd/tools/vtest-parser.v (to get more sensible backtraces),
then: v test-parser x_fuzz.v .

Thanks @joe-conigliaro for reporting the problem.

[joe-conigliaro]

Thanks @spytheman, excellent work!