Merge #527

527: goto defenition works for type-inferred methods r=flodiebold a=matklad

This uses type inference results for `goto method` functionality.

This is achieved by adding another map to `InferenceResult`. I wonder how we should handle this long-term... The pattern seems to be "we are doing some analysis, and we produce some stuff as a by-product, and IDE would like to use the stuff". Ideally, adding an additional bit of info shouldn't require threading it through all data structures.

I kinda like how Kotlin deals with this problem. They have this [`BindingContext`](https://github.com/JetBrains/kotlin/blob/72e351a0e3610051fe4222dca4e1eeedf7ae45da/compiler/frontend/src/org/jetbrains/kotlin/resolve/BindingContext.java#L122) thing, which is basically an [`AnyMap`](https://github.com/JetBrains/kotlin/blob/72e351a0e3610051fe4222dca4e1eeedf7ae45da/compiler/frontend/src/org/jetbrains/kotlin/resolve/BindingContext.java#L122) of HashMaps.

Deep in the compiler guts, they [record the info](https://github.com/JetBrains/kotlin/blob/ba6da7c40a6cc502508faf6e04fa105b96bc7777/compiler/frontend/src/org/jetbrains/kotlin/resolve/calls/tasks/TracingStrategyForInvoke.java#L70-L75) into the map, using a type key, a value key and a value.

Then the IDE [reads this map](https://github.com/JetBrains/kotlin/blob/ba6da7c40a6cc502508faf6e04fa105b96bc7777/idea/src/org/jetbrains/kotlin/idea/inspections/RedundantNotNullExtensionReceiverOfInlineInspection.kt#L64) (via a [helper](https://github.com/JetBrains/kotlin/blob/ba6da7c40a6cc502508faf6e04fa105b96bc7777/compiler/frontend/src/org/jetbrains/kotlin/resolve/calls/util/callUtil.kt#L178-L180)). The stuff in between does not know that this type-key exists, unless it inspects it.

Co-authored-by: Aleksey Kladov <aleksey.kladov@gmail.com>