Respect original column type when inserting

[kasiafi]

In JDBC connectors, currently the insert target type for a column depends
on the Presto type chosen to represent the column's values.
This strategy doesn't support multiple original types mapped to the same
Presto type -- a single WriteMapping imposes a target type which not necessarily
fits them all, and so inserts may cause errors at finish time when trying to copy
the temporary table into the target table (unless the target database is able
to perform a coercion).
This is fixed by making INSERT aware of the target column types i.e. by using
the original column type in the temporary table and obtaining the write function
from the ColumnMapping.

This will allow to support inserting Postgres array values when they are mapped to Presto JSON (issue #682 addressed by #1148)

Fixes #338

[findepi]

The JdbcPageSink is JDBC-specifc, and so JdbcOutputTableHandle is.
It's surprising at first that the argument isn't cast by the caller but it's being cast here.

I would consider something like this: pass JdbcOutputTableHandle (as it was).
Inside constructor, check:

if (!handle.getJdbcColumnTypes().isPresent()) {
    ... the old way
}
else {
    ....  the new way
}

(This would be further simplified if we combine WriteFunction and WriteNullFunction into a single interface, but let's not do this in this PR.)

[kasiafi]

Applied comments. Travis fails repeatedly due to out of memory error.

[findepi]

Merged, thanks!

[pratyakshsharma]

@findepi @kasiafi I was trying to understand the changes in this PR. Referring to this statement - This strategy doesn't support multiple original types mapped to the same Presto type -- a single WriteMapping imposes a target type which not necessarily fits them all, and so inserts may cause errors at finish time when trying to copy the temporary table into the target table (unless the target database is able to perform a coercion)., can you give an example of the scenario or data type where the inserts would fail? Thanks!

[findepi]

answered here: https://trinodb.slack.com/archives/CP1MUNEUX/p1647499864386799