Describe explicit minute 0 when seconds are not implicit

[austek]

Description of your changes

TimeDescriptionStrategy always suppressed the minute when it was On(0), treating minute 0 as an implied default. That reads well when the seconds are implicit too, but when they are not, the description asserts a sub-minute frequency with nothing bounding it to the minute the cron actually fires on:

* 0 9-23 * * ?   before: every second every hour between 9 and 23
                 after:  every second at 0 minute every hour between 9 and 23

This is the case reported in #3 by @natrajmolala, where you invited a PR with a test.

The fix gates the existing suppression on the seconds also being default. Three related expressions were wrong in the same way and are corrected by the same condition:

expression	before	after
5 0 9-23 * * ?	at 5 second every hour between 9 and 23	at 5 second at 0 minute every hour between 9 and 23
0/5 0 * * * ?	every 5 seconds from second 0	every 5 seconds from second 0 at 0 minute
* 0 */4 * * ?	every second every 4 hours	every second at 0 minute every 4 hours

Where the suppression is deliberate it is preserved — 0 0 9-23 * * ? still describes as "every hour between 9 and 23" and 0 0 10 * * ? as "at 10:00". Issue3Test pins those too, so the collapse cannot regress.

Scope

#3 is an umbrella issue and this closes one reported case out of four; the others are phrasing rather than correctness, so the commit says Refs #3 rather than Fixes.

Incidental

The strategy set is now insertion ordered and the duplicated predicate call in the match loop is hoisted. Both are hardening only — the eleven registered predicates cover mutually exclusive (hours, minutes, seconds) type triples, so the previous HashSet iteration order was not reachable as nondeterministic output.

Impact Analysis

Output changes only for expressions that fire on an explicit minute 0 with non-default seconds, which were describing roughly sixty times the firings they actually produce. Callers asserting on those exact strings will need updating. No API, parser or i18n bundle changes, so no new resource keys and no impact on the non-English locales.

Checklist

• I have performed a self-review of my own code
• My changes generate no new warnings
• Existing tests pass: 810 tests, 0 failures, 0 errors (up from 802 on master by the 8 added cases; 57 pre-existing skips)