Time-travel & Incremental API to support more timestamp formats

[xushiyan]

Description of the improvement

Currently Table::read_incremental_records() expects start and end timestamps to be in the Hudi timeline format, i.e., yyyyMMddHHmmSSSSS or the legacy yyyyMMddHHmmSS. We should support parsing more timestamp formats.

Sames goes for Table::read_snapshot_as_of() and other *as_of() APIs.

Expected behavior

• To support parsing strings in the form of epoch time (including seconds, milliseconds, microseconds, nanoseconds) and convert to Hudi timeline format for further processing.

• To support parsing strings in the form of ISO8601 format like

  • 2019-01-23T12:34:56.123456789+00:00
  • 2019-01-23T12:34:56.123456+00:00
  • 2019-01-23T12:34:56.123+00:00
  • 2019-01-23T12:34:56+00:00
  • 2019-01-23T12:34:56Z and other precisions like above
  • 2019-01-23

Additional context

The conversion to Hudi timeline format should consider the hudi table's timeline timezone config.

[hanbings]

hi, I'm mildly interested in this issue and would be open to giving it a try if possible.