Report an error for over-large repetition counts instead of a literal (#663)

[xhon-pelushi]

Problem

a{1000000000} parses successfully (as the literal string a{1000000000}), while smaller over-limit counts like a{10000} correctly fail with "invalid repetition size" (#663). The maximum repetition count is 1000, so both should error.

ParseInteger() (re2/parse.cc) bails out mid-number to avoid overflow:

while (digit) {
    if (n >= 100000000)
        return false;        // stops before consuming all the digits
    n = n*10 + c - '0';
    ...
}

Returning false makes MaybeParseRepetition() fail, so { is treated as a literal instead of a repetition — the count never reaches the maximum_repeat_count check in PushRepetition().

Fix

Keep consuming the digits and cap the accumulated value instead of bailing out, so the repetition is still recognized and then rejected as too large:

while (digit) {
    if (n < 100000000)
        n = n*10 + c - '0';  // cap (avoids overflow), keep consuming
    ...
}

n is bounded at < 1e9 (well within int), and PushRepetition() rejects any min/max > 1000 with kRegexpRepeatSize, so a{1000000000} now reports "invalid repetition size" like the other over-limit counts. Values at or below the limit are unaffected.

Test

Adds a{1000000000} and a{0,1000000000} (high position of the range) to badtests[] in re2/testing/parse_test.cc, which asserts Regexp::Parse(...) == NULL. Both currently parse as literals (so the test fails on main) and error with this change.

Fixes #663