Add Decimal128 support

[gagik]

This PR add support for Decimal128 as fields and in RealmAny.

class Decimal128Object: RealmObject {
    val decimal128Field: Decimal128 = Decimal128("1.23422354E53")

    val any: RealmAny = RealmAny.create(Decimal128("1.234523E34"))
}

TODO:

• Update to core containing [C-API] Querying on Decimal128 field with value NaN does not find objects realm-core#6182

Closes #653

[gagik]

Here I had to specify org.mongodb.kbson.BsonDecimal128 as it would error with org.mongodb.kbson.Decimal128. Seemed to work fine regardless but not too sure if this would cause errors in some cases.

[rorbech]

The core fix for querying on NaN is merged, but it is done on top of Core 13 so need that to remove that last FIXME in the roundtripSpecialValues ... but should be up for an initial review.

[edualonso]

Looks good but I think you are missing testing Decimal128 inside RealmAny values in lists and sets and by extension aggregations in QueryTests. You should add a RealmAny.create(Decimal128(...)) entry in

realm-kotlin/packages/test-base/src/androidAndroidTest/kotlin/io/realm/kotlin/test/shared/RealmListTests.kt

Line 1409 in 0ffda2f

internal val REALM_ANY_PRIMITIVE_VALUES = listOf(

[edualonso]

Nice improvements, especially to the descriptors and the new flags. Also worth noting your comment during our conversation about moving TypeDescriptor.aggregateClassifiers and TypeDescriptor.anyClassifiers, so for now keep the changes to a minimum. But overall looks great.

[cmelchior]

Nice 🚀 Only a few minor questions/comments.

[cmelchior]

This is probably correct, but should this not be ulongs?

[rorbech]

JNI does not support ULong so need to assign it as a long array. But updated to use the experimental unsigned implementation of

w = ulongArrayOf(it.low, it.high).toLongArray()

This is symmetric to what we are doing on the output side in https://github.com/realm/realm-kotlin/pull/1179/files#diff-ecf6dda367cfda31a14c17876fa5eacac9bf99b8a34f9c9b71d6da4b8d062c96R45.

The values are passed to the C-API as the signed reinterpretation of the corresponding elements as described in https://kotlinlang.org/api/latest/jvm/stdlib/kotlin.collections/to-long-array.html

Returns an array of type LongArray, which is a copy of this array where each element is a signed reinterpretation of the corresponding element of this array.

And the C-API does the similar unsigned reinterpretation on the other side. I have verified the special values written in https://github.com/realm/realm-kotlin/pull/1179/files#diff-e3a1343b699de7bd2efcc1bd08f8f0406dc545438996074472ce7dbcd36eb159R35 is actually displayed correctly when opening the written realm with Realm studio.

[cmelchior]

It feels a bit weird that MINMAX covers two different ways. It would probably make more sense to split them?

[rorbech]

I don't think we would ever have a type that supports min but not max, but now I split it. It is also just in test infrastructure code so not that critical.