Retrying Job Scheduler

Problem

Design RetryingJobScheduler to schedule one-time jobs and retry failed execution attempts. A claim identifies one execution attempt for a job.

Each job moves from waiting to running when claimed. A successful result ends its lifecycle, while a failed result gives the same job a new due time.

Requirements

Registration. schedule creates a waiting job with dueAt equal to runAt and attemptNumber 0.
Eligibility. A job is eligible exactly when it is waiting for its first attempt or a retry and dueAt <= now; returned claim order is unspecified.
Claim attempts. claimDue starts every eligible job, increments its attemptNumber, and returns one Claim containing its jobId and new attempt number. A running job is not eligible.
Failure. Reporting failure keeps the same job and attempt number, then makes the job eligible again at now + retryDelaySeconds; only the next claim increments the attempt number.
Success. Reporting success finishes the one-time job permanently, so it is never eligible again.

API

Signature	Returns	Behavior
`RetryingJobScheduler(retryDelaySeconds: integer)`	A new `RetryingJobScheduler`	Creates an empty scheduler with one fixed positive retry delay.
`schedule(jobId: string, runAt: integer)`	`void`	Registers a one-time job.
`claimDue(now: integer)`	`Claim[]`	Starts attempts for all eligible jobs and returns their claims.
`reportResult(jobId: string, succeeded: boolean, now: integer)`	`void`	Resolves the job's active attempt.

Claim has fields jobId: string and attemptNumber: integer.

Examples

Given RetryingJobScheduler(5):

Call	Result
`schedule("email", 10)`	No return value
`claimDue(9)`	`[]`
`claimDue(10)`	`[Claim("email", 1)]`
`claimDue(11)`	`[]`
`reportResult("email", false, 12)`	No return value
`claimDue(16)`	`[]`
`claimDue(17)`	`[Claim("email", 2)]`
`reportResult("email", true, 18)`	No return value
`claimDue(100)`	`[]`

Constraints

1 <= retryDelaySeconds <= 1,000,000.
0 <= runAt, now, dueAt <= 9,000,000,000,000,000.
jobId is nonempty, has at most 64 characters, and is compared by exact equality.
Each jobId is scheduled exactly once.
At most 1,000 jobs and 10,000 method calls occur in one testcase.
reportResult names a running job.
Later now values do not decrease.
Calls outside these constraints are not judged.

Notes

All calls are sequential. Concurrency, recurring schedules, cancellation, workers, persistence, and distributed coordination are outside the contract.