[Feature Request] Improve success rate of a rerun by splitting previously failed partitions into sub-partitions using a multiplier

**Issue:** When we have a few token-ranges that fail during a CDM run, many of them tend to fail on reruns as well as the partitions in the token-range may be too large or have some other similar issues. **Fix:** Allow breaking a token-ranges into smaller token-ranges during a rerun. **Note:** This options is to be used only when other CDM options (e.g. batchAize, fetchSize, ratelimit) do not help reduce errors. **Context:** This is a common issue when migrating tables with large partitions or tables with many uneven partition sizes.