Support canonicalized closures

[DartBot]

This issue was originally filed by @seaneagan

---

If a closure only captures scope names which are compile-time-constant, then it itself is compile-time-constant, and thus should always evaluate to the same exact closure object. This should be relatively easy to implement, since it can be accomplished by the compiler itself.

If a closure captures the same non-compile-time-constant variables on multiple evaluations, then each such evaluation could produce the same closure object. This would presumably require storing a table to map between captured variables, and the resulting closures. Thus, this might be a bit more difficult to implement, at least in JavaScript.

[DartBot]

This comment was originally written by @seaneagan

---

Also, method closurization should be idempotent. This falls under the category of the second paragraph above since this is not compile-time-constant, however, it is a less general case.

[anders-sandholm]

Added Area-Language, Triaged labels.

[gbracha]

Set owner to @gbracha.
Added Accepted label.

[gbracha]

Issue #2402 has been merged into this issue.

[anders-sandholm]

Added this to the M1 milestone.

[gbracha]

Somewhat related to 144.

[gbracha]

see also http://code.google.com/p/dart/issues/detail?id=3060 and http://code.google.com/p/dart/issues/detail?id=144. All these should be considered and resolved in a cohesive fashion.

[gbracha]

Static methods and top level functions are compile-time constants starting with the 0.10 draft. However, we have not taken any such decision on arbitrary functions.

---

Removed Type-Defect label.
Added Type-Enhancement, Done labels.

[gbracha]

Analyzer issue #3404
dart2js issue #3405
VM issue #3406