fix: fix getLong on NUMERIC

[rayudu3745]

Updated the getLong getter in JdbcResultSet for GoogleSQL NUMERIC columns to use spanner.getBigDecimal(spannerIndex) instead of getString(). This prevents an IllegalStateException generated by the underlying Spanner library

we call .toBigInteger() on the retrieved BigDecimal value to truncate fractional parts (e.g., 3.14 becomes 3).

Other integer getters (getByte, getShort, getInt) in JdbcResultSet and generic type conversions in JdbcTypeConverter have been updated accordingly

[olavloite]

So one interesting thing regarding NaN is that the following assertion actually passes:

assertEquals(0, (int) Float.NaN)

Do you know what the PostgreSQL JDBC driver returns in a case like this:

1. Does it return zero (which would be consistent with how Java casts a NaN to int)?
2. Or does it throw an error like here?

[rayudu3745]

verified pgdriver behaviour, it also throws exception for "NaN" for these methods

[olavloite]

nit: can we use assertThrows(...) for these assertions instead of this try-fail construct. I know that there are many of them in this project, but that is because the original JDBC driver was written when Java 7 was the minimum supported version, which did not support assertThrows(..) (as it did not support lambda expressions)

[rayudu3745]

done

[olavloite]

Is this really needed? I don't remember having seen any issues with this in the past for this repository. But it might be that recent versions of the emulator are more sensitive to this. (Could we maybe make it a bit more dynamic than always sleeping for 1.5 seconds?)

[rayudu3745]

done