reject an empty quoted abbreviation in ParseAbbr

[rajath201]

ParseAbbr() bounds the length of the unquoted abbreviation form but not the <...> one, so <> parses as a zero-length abbreviation. An empty dst_abbr is how PosixTimeZone signals "no daylight time" (time_zone_posix.h, and the if (posix.dst_abbr.empty()) branch in ExtendTransitions()), so a footer like EST5<>,M3.2.0,M11.1.0 loads with its DST rule quietly dropped and load_time_zone() still returns true. Looking up 2035-07-01T12:00:00Z gives -18000/EST against 2 transitions, where EST5EDT,M3.2.0,M11.1.0 gives -14400/EDT against 804; <>5 does the same to std_abbr and leaves %Z empty.

Add the matching length check to the quoted branch, beside the existing p - op < 3, so the constraint lives where the abbreviation is built instead of asking each consumer to treat empty as two different things. tzcode rejects the same input in tzparse() (if (!stdlen) return false;) and RFC 9636 wants at least one character between the brackets. No valid zone shifts: all 485 under testdata/zoneinfo give byte-identical descriptions, lookups, offsets and abbreviations before and after.

[devbww]

Given that we know p - op >= 1 at this point, I would weaken the conditional to:

  if (p - op == 1) return nullptr;  // no "<>"

This also makes it very clear that the following size, (p - op) - 1, is not zero.