test: separate scalars test for D1 with QC (#5598)
[ORM-1357](https://linear.app/prisma-company/issue/ORM-1357/fix-d1-data-types-issue)
This PR separates the D1 scalars test between QE and QC.
QE currently returns the type of the first row as the column type for
numeric values and performs no attempts to ensure that the numbers in
that column conform to the inferred type. We don't really want this
behavior going into the future, so I separated the test to have a
different expectation when running with QC.
When QC is used, we return the column type of numeric values as
`unknown` and don't attempt to convert them. This affects the `types`
returned from raw queries. In some cases this can also affect
serialization of rows - when QE determines a column to be a BigInt, it'd
get serialized as a string, but since this detection is not really
reliable (it only looks at one row) and the actual numbers come out as
javascript numbers from the database, it seems entirely pointless, so we
do no equivalent conversion.
