Add support for DecimalType in Remainder for Spark 3.4 and DB 11.3 [databricks]

[NVnavkumar]

Fixes #8161.

This reverts the changes to integration tests made in #7609, and adds support for Remainder with DecimalType in Spark 3.4 and Databricks 11.3.

This does this by casting the operands using the following formula:

scale = max(s1, s2)
precision = max(p1-s1,p2-p2) + scale
lhsType = rhsType = DecimalType(precision, scale)

This enables just enough precision and scale to compute the remainder. This gives enough scale to produce the part of the output that is < 1, and enough precision to contain the values of both operands.

The output type is actually this formula:

scale = max(s1, s2)
precision = min(p1-s1,p2-p2) + scale
outputType = DecimalType(precision, scale)

Which is only just enough precision and scale to store the remainder.

[NVnavkumar]

build

[NVnavkumar]

build

[revans2]

Feel free to ignore the nit if it means we can be faster at getting the full fix in by 23.06.

[revans2]

nit: it would be nice to add in a psNote for DECIMAL128 here to explain what we don't fully support. This is really minor if we think we can get full remainder support in before we ship 23.06. (especially minor because we don't generate the support docs for anything but the oldest version of Spark that we support) Oh well....

[revans2]

Looks like you have some lines over 100 chars that need to be fixed.

[NVnavkumar]

build

[revans2]

build

[NVnavkumar]

Follow up for Decimal128 support is here: #8330

[gerashegalov]

Suggested change

	if (a.left.dataType.isInstanceOf[DecimalType] &&
	a.right.dataType.isInstanceOf[DecimalType]) {
	val lhsType = a.left.dataType.asInstanceOf[DecimalType]
	val rhsType = a.right.dataType.asInstanceOf[DecimalType]
	val needed = DecimalRemainderChecks.neededPrecision(lhsType, rhsType)
	if (needed > DType.DECIMAL128_MAX_PRECISION) {
	willNotWorkOnGpu(s"needed intermediate precision ($needed) will overflow " +
	s"outside of the maximum available decimal128 precision")
	}
	}
	(a.left.dataType, a.right.dataType) match {
	case (lhsType: DecimalType, rhsType: DecimalType) =>
	val needed = DecimalRemainderChecks.neededPrecision(lhsType, rhsType)
	if (needed > DType.DECIMAL128_MAX_PRECISION) {
	willNotWorkOnGpu(s"needed intermediate precision ($needed) will overflow " +
	s"outside of the maximum available decimal128 precision")
	}
	case _ => ()
	}