…se (#23741)

* Add CronExpression and CronScheduler utilities to datadog_checks_base

CronExpression parses a standard 5-field UTC cron string and computes the
next/previous scheduled tick relative to any timestamp. CronScheduler wraps
an expression with a non-blocking 'has my tick elapsed?' state machine
designed to be polled once per AgentCheck.check() invocation, with an
optional startup_lookback for recovering a missed tick across check
restarts.

Supports *, n, n-m, n,m,k, */k, and n-m/k in each field, plus 7 as a
Sunday alias for day-of-week. Vixie OR semantics for day-of-month vs.
day-of-week when both are restricted.

* Add changelog entry for cron utility

* Fix N/step cron semantics, tighten error types, derive equality from parsed fields

- N/step in a single-integer base now extends to the field's high (Vixie
  semantics): "5/15 * * * *" fires {5, 20, 35, 50} instead of {5}.
- CronExpression raises TypeError for non-string input and ValueError for
  all malformed expressions, including parse-time rejection of DOM/month
  combos that can never fire ("0 0 31 2 *").
- Equality and hashing derive from the parsed fields, so "0 0 * * 7" and
  "0 0 * * 0" compare equal.
- _jump_month iterates self._months directly to match the other _jump_*
  helpers; the _roll_month helper is removed.
- CronScheduler.due_ticks returns list[float] matching the materialised
  behaviour; the typing.Iterator alias drops with it.
- Rename test helpers to their fully spelled form (utc_timestamp,
  iso_format, BASE_TIMESTAMP); drop em-dashes and motivational framing
  from comments and docstrings.

* Promote MAX_DAYS_PER_MONTH to module scope and tighten DOM/month guard

- Move the per-month max-day table from inside _reject_unsatisfiable_dom_month
  to a module-level MAX_DAYS_PER_MONTH constant, alongside the other range
  constants.
- Short-circuit _reject_unsatisfiable_dom_month when DOM is unrestricted as
  well: only the DOM-restricted + DOW-unrestricted case can be impossible at
  parse time.
- Add a NEXT_PREV_CASES row that crosses a month boundary backward while
  Vixie OR semantics are active ("0 0 31 1 3" anchored in Feb 2026), pinning
  the interaction between _jump_month and _day_matches.
- Drop the now-redundant list(...) wrappers around CronScheduler.due_ticks
  calls in the scheduler tests; due_ticks already returns list[float].

* Split _parse_field into single-purpose helpers and align backward jumps

- _parse_field now returns only the sorted values tuple; the is_restricted
  flag was discarded by three of five call sites. CronExpression.__init__
  computes _dom_restricted and _dow_restricted directly from the raw fields
  where the Vixie OR rule actually needs them.
- _parse_field delegates to four helpers, each with a single responsibility:
    _expand_chunk: one comma piece, with optional /step
    _split_step:   extracts and validates /step
    _parse_base:   dispatches between '*', 'n-m', and 'n'
    _parse_range:  parses an 'n-m' inclusive range
  Each helper raises with its own context so error messages still pinpoint
  the specific malformation.
- _jump_hour and _jump_minute backward paths now use self._minutes[-1]
  (and self._hours[-1]) instead of hardcoded 59/23 when crossing a
  boundary, so a sparse-minute schedule does not need an extra walk
  iteration to recover.

* Add edge-case tests adapted from croniter's test suite

Covers behavior surface our parser claims to support but had no dedicated
test:

- Duplicate and overlapping values in comma lists normalize to the same
  parsed form (e.g. "5,5,1-6 * * * *" equals "1-6 * * * *").
- Vixie OR with stepped DOM and a weekday ("0 16 */2 * 6"), pinning the
  interaction between */k expansion in DOM and DOW match in _day_matches.
- next_tick and previous_tick around stepped DOM ("5 0 */2 * *") from an
  even-numbered day where the anchor itself does not match.
- previous_tick crossing into the previous month with a simple DOM
  ("0 0 22 * *" anchored mid-March -> 22 February).
- Weekend DOW list ("0 0 * * 0,6") matching the most recent Sat/Sun in
  both directions.
- Comma list in the month field ("0 0 1 1,3,6,9,12 *") rolling forward
  and backward across allowed months.

* Reject cron expressions with whitespace inside fields

A field like "1, 1, 3,3" looks at a glance like one comma list, but the
spaces split it into multiple whitespace-separated fields, so the overall
expression no longer has five fields. The parser already rejects this
correctly via the field-count check; pinning the behaviour here covers
the common ways a user might accidentally introduce intra-field
whitespace (after a comma, before a comma, around a dash).

* Accept whitespace around commas in cron fields

A field like "1, 2, 3" is visually unambiguous: the writer means a
comma list. The previous parser was strict and rejected it as
multiple whitespace-separated fields. Normalise whitespace adjacent
to commas before splitting into fields, so "0 0 1 1, 2, 3 *" parses
the same as "0 0 1 1,2,3 *".

Dashes stay strict ("1- 3" is more likely a typo than a style
choice, so it is still rejected).

* Reject any whitespace inside a cron field

The previous commit accepted whitespace around commas but not around
dashes, which is inconsistent: every kind of intra-field whitespace
has only one plausible reading (the punctuation it sits next to), so
"unambiguous to a human" applies the same way to "1, 2" and "1- 3".
A single strict rule (no whitespace inside fields) gives the user
the same clear failure for every variant and matches standard cron,
croniter, and systemd timers.

* Slim MAX_DAYS_PER_MONTH to non-31 months and merge tick-table tests

- Drop the always-31 entries from MAX_DAYS_PER_MONTH and look them up via
  .get(month, 31) so the table communicates "31 is the rule, these are
  the exceptions." Add a short comment explaining why February uses 29.
- Collapse test_next_tick and test_previous_tick into a single
  test_bracketing_ticks with the cases inlined into the parametrize block.
  Removes the top-level NEXT_PREV_CASES constant and the underscore-prefixed
  unused arguments, while preserving the joint (prev, next) specification at
  each anchor. Failure messages now name the broken direction.

* Share a single parser between construction and is_valid

- Extract _parse() as the one grammar both __init__ and is_valid use, so
  is_valid validates without instantiating a CronExpression.
- Explain the scheduling sense of "due" in the due_ticks docstring.

* Drop is_valid; document walk budget and CronExpression exceptions

- Remove the is_valid predicate and update tests to validate through the
  constructor.
- Add a comment explaining WALK_ITERATION_BUDGET.
- Document the ValueError and TypeError CronExpression raises on bad input.

* [Release] Bumped datadog_checks_base version to 37.38.0

* [Release] Update metadata

* Revert "Aarakke/release checks base (#23877)"

This reverts commit 8abadad.

* Restore in-toto from master

* [Release] Bumped datadog_checks_base version to 37.39.0

* [Release] Update metadata