Computing the Next Run Time

next_run(cron, after) returns the smallest datetime > after that matches the cron expression.

A naive implementation increments by 1 minute and tests matches() each step, but that's slow for 0 0 1 1 * (yearly). The standard algorithm walks field-by-field, smallest-to-largest, snapping each component up to the next allowed value:

1. Increment minute to next allowed (carry if needed).
2. If minute carried, increment hour. If hour > max-allowed, snap to next allowed
   hour and reset minute to its smallest allowed value.
3. Similar for day, month, year.
4. Verify matches(); if not (e.g. day-of-week mismatch), advance by a day and retry.

Edge case: years with no matching dates (e.g. * * 30 2 *) should not loop forever — cap at +4 years and return error.