Fix converting rows with FLOAT4 and BOOLEAN

[iwysiu]

From investigating with our external dev credentials, it looks like we were converting Float4 and Boolean from the wrong types. When I set the redshift column to the REAL type, it returned a Float4, so I don't think our issue was that we weren't supporting the type. Also, I suspect that the FLOAT type should also be converted as a double, but in Redshift it's synonymous with DOUBLE and I couldn't get it to actually return that type.
I'm a little suspicious of the solution because it implies either that FLOAT4 has never worked or that AWS changed the return type.

Fixes #204

[github-actions]

Backend code coverage report for PR #213

Plugin	Main	PR	Difference
redshift	24.0%	24.0%	0%

[github-actions]

Frontend code coverage report for PR #213
No changes

[fridgepoet]

Nice finds!

Do you think we should just consider FLOAT the same case as FLOAT8? It's what I'm seeing for the DOUBLE PRECISION column type.

I'm a little suspicious of the solution because it implies either that FLOAT4 has never worked or that AWS changed the return type.

Same about the boolean, right?

By the way, for future reference if we want to test this out, it's much easier to write a query like SELECT CAST('123456' AS REAL); than to create a cluster/table/column of a certain type.

[fridgepoet]

Just wondering out loud: Regarding all the special checks for time column (#62), I wonder why we don't ever do it for other "number" types. Or why we don't somehow ensure it is PostgresSQL before just going for the assumption it is Unix time.