Support indexes on all compatible types

[nielsenko]

Allow (and correctly apply) @Indexed() on int, bool, String, ObjectId, Uuid, and DateTime (and the nullable version as well).

@RealmModel()
class _Indexable {
  @Indexed()
  late bool aBool;
  @Indexed()
  bool? aNullableBool;
  @Indexed()
  late int anInt;
  @Indexed()
  int? aNullableInt;
  @Indexed()
  late String aString;
  @Indexed()
  String? aNullableString;
  @Indexed()
  late ObjectId anObjectId;
  @Indexed()
  ObjectId? aNullableObjectId;
  @Indexed()
  late Uuid anUuid;
  @Indexed()
  Uuid? aNullableUuid;
  @Indexed()
  late DateTime aDateTime;
  @Indexed()
  DateTime? aNullableDateTime;
}

Fix #797,

[nirinchev]

Not sure if it's worth writing a test for this, but we could try to prove that the index exists in Core and is being used by opening 2 realms - one with indexed fields, the other with non-indexed fields, then adding 10000 objects and querying for a single object. The test should expect that the query utilizing the index completes much faster.

[nielsenko]

Not sure if it's worth writing a test for this, but we could try to prove that the index exists in Core and is being used by opening 2 realms - one with indexed fields, the other with non-indexed fields, then adding 10000 objects and querying for a single object. The test should expect that the query utilizing the index completes much faster.

Added test of 1.000 lookups against 100.000 objects and compared indexed vs not-indexed for all indexable types, and assert the speedup is at least 10 fold.

[blagoev]

Its good, I have couple of questions.

[nirinchev]

Small nits for the tests, otherwise looks good.

[nirinchev]

Why not

Suggested change

	final searchOrder = (List.generate(max, (i) => i)..shuffle(Random(42))).take(1000).toList();
	final random = Random(42);
	final searchOrder = List.generate(1000, i => random.nextInt(max));

I know perf doesn't matter that much, it's just a little easier to read.

[nielsenko]

Problem is it can create duplicates. That is what I avoid above

[nirinchev]

It can, but do we care?

[nirinchev]

I think we should be stricter here - something like expect(indexedTime, lessThan(0.5 * notIndexedTime))

[nielsenko]

I used to have

expect(indexedTime * 10, lessThan(notIndexedTime));

but I was asked to loosen it

[nielsenko]

.. but maybe 2 would be fine 😄