[FEA] non-UTC time zone feature needs a new kernel to parse time string in a local time zone.

[res-life]

Is your feature request related to a problem? Please describe.
For casting string in a specific time zone to timestamp, Spark first parse string to segments(year, mohth, day, hour, minute, seconds ...), then create a local time from the segments, then convert the local time in the specific time zone to UTC time. Refer to: spark code .
Currently, TimeZoneDB APIs are:

• fromTimestampToUtcTimestamp(timestamp_column(long cv), zoneId)
• fromUtcTimestampToTimestamp(timestamp_column(long cv), zoneId)

They are not enough, we should implment a new API like:

• parseLocalTimestampString(string_column, zoneId)

Describe the solution you'd like
In parseLocalTimestampString kernel
First parse string to segments, refer to this JNI PR, then transform segments in the specific time zone to UTC time. The transform needs to query the time zone transition/transition rule tables. Kernel should provide:

• def transform_from_local_time_segments_to_utc(zongId, int year, int month, ... int nanoSeconds) : Long

So the overall logic will be:

• parse string to segments(year, month, day ...)
• transfrom segments to UTC timestamp long value

Note:
-Parser should provide a option to to handle special string: EPOCH, NOW, ...... for Spark before 320.

• For Ansi mode, should investigate how to handle this mode efficiently.

[res-life]

Similar to this issue: #6846

[res-life]

Close this because it's tracked by #6846