FTP: More EPLF 'm' timestamp parsing fixes

[MegaManSec]

The m case in ftpListParseParts() reversed the strtol() validity
check. We broke out when any digits were consumed instead of only when
no conversion happened. As a result, valid epoch timestamps were treated
as invalid and p->date was often left unset.

Also guard against ctime() returning nil and avoid null pointer
dereference if no newline is present in ctime() output.

Also reject m entries with trailing garbage.

[yadij]

"good enough". Long-term we need to replace the parsing with a Tokenizer.

[rousskov]

@eduard-bagdasaryan, please run this PR through the tests that you did for recent commit 965e255. I am a bit surprised that the reversal of that tmp condition has not affected those tests; I am not claiming that the official condition was correct.

[eduard-bagdasaryan]

run this PR through the tests

Done, no problems were found.

My primary concerns have been addressed.

[rousskov]

I adjusted PR description a little to mention trailing garbage handling and to avoid retelling the fairly clear diff when it comes to strstr() refactoring. I am not against adding some text about that refactoring though.

[rousskov]

Alex: I wonder whether whitespace ought to be allowed after the mNNN value.

Francesco: A quick reading of https://cr.yp.to/ftp/list/eplf.html says no, it ought not be allowed

Agreed. However, the same document also prohibits NUL after the mNNN value -- it requires a comma, even if m is the last fact in the list of facts. Thus, AFAICT, this code should become something like this:

Suggested change

	if (*secondsEnd && *secondsEnd != ',')
	break; // trailing garbage after the seconds value
	if (!secondsEnd || *secondsEnd != ',')
	break; // missing comma terminator after the `m` fact value

Ideally, this check should be placed after seconds validation currently located below.