Add support for string based/dynamic API on RealmObject

[nirinchev]

Adds a string-based API to read properties of a RealmObject. Right now we only support reading as that's required to get migrations working, but it can be extended in the future to also support setting.

final obj = realm.dynamic.find("Foo", 123);

// Get simple primitive
final str = obj.dynamic.get<String>("stringProp");

// Get another object
final bar = obj.dynamic.get<RealmObject>("bar");
final barString = bar.dynamic.get<String>("stringProp");

// Get a list of objects
final bars = obj.dynamic.getList<RealmObject>("bars");

// Iterate the list as normal
for (final bar in bars) {
  print(bar.dynamic.get<DateTime>("dateProp");
}

Additionally, implements noSuchMethod on RealmObject in order to support dynamic invocation:

dynamic obj = realm.dynamic.find("Foo", 123);

// Get simple primitive
String str = obj.stringProp;

// Get another object
dynamic bar = obj.bar;
String barString = bar.stringProp;

// Get a list of objects
final bars = obj.bars;

// Iterate the list as normal
for (final bar in bars) {
  print(bar.dateProp);
}

Part of #70.

Note: This only adds support for reading properties, but not for setting them. It won't be difficult to extend the API and support setting of properties, but is out of scope for this PR.

[nirinchev]

This liberalizes the type args on RealmList to support nullable objects. It doesn't actually add support for lists of nullables - that is done in #708, but is necessary to simplify a lot of the generic constraints elsewhere.

[nirinchev]

This must be a change with the new ffi and seems like it was a bug - not sure how/why our tests didn't catch it, but likely warrants an investigation.

[blagoev]

added a test and fixed it in #854. This can be left here as well

[nirinchev]

The changes in this file are mostly whitespace due to grouping tests together. I recommend reviewing with ?w=1.

[blagoev]

added a test and fixed it in #854. This can be left here as well

[blagoev]

I suggest we don't hide this under our own name

[nirinchev]

This is just a typedef - it's not going to affect users. The reason why I need to typedef it is to avoid a collision with the dynamic property on RealmObject.

[nielsenko]

Won't users be hit by the same?

[blagoev]

I guess we can add more tests for the new validations