source-zendesk-support: checkpoint during AuditLogs stream

[Alex-Bair]

Description:

For tasks that have a significant number of audit logs in Zendesk Support, it can take more than 24 hours to backfill. With the current AuditLogs stream strategy of reading records in descending order then stopping when we've reached a record we already read, we can't checkpoint until the backfill is completed. This means tasks that take more than 24 hours to backfill AuditLogs can get stuck restarting & attempting to backfill.

This PR updates the AuditLogs stream to read records in ascending order. This lets the connector checkpoint after each response & pick up where it left off when it's restarted.
Some noteable changes include:

• Passing a list containing start_time and end_time for the filter[created_at][] query param to bound the responses' results to the specified timespan.
• Pushing end_time 30 seconds in the past to avoid missing records if we query Zendesk in the middle of creating multiple records with the same created_at.
• Keeping AuditLogs on the older get_updated_state method of updating state instead of migrating to the newer state property. This is to avoid rebackfilling all existing tasks.

Workflow steps:

(How does one use this feature, and how has it changed)

Documentation links affected:

(list any documentation links that you created, or existing ones that you've identified as needing updates, along with a brief description)

Notes for reviewers:

The upper time bound for the AuditLogs stream is moved a little in the past to avoid potential issues where we query Zendesk Support when its in the middle of creating multiple records with the same created_at. I don't have evidence that this can actually occur, but it sounds possible to me and a 30 second delay is relatively insignificant when the connector is running on 5 minute intervals.

Existing tasks do not need backfilled. The AuditLogs cursor value has remained the same (created_at for the most recent record) and the way the connector manages this state is the same.

Tested on a local stack. Confirmed:

• Audit log records are received in ascending order.
• State is checkpointed & documents are emitted after each response.
• Checkpointed AuditLogs state is used after restarting the connector during a backfill.

---

This change is 

[williamhbaker]

LGTM