Preserve directly deconstructed tuple values in analyzer flow

[jtschuster]

Summary

In the analyzer, we see ITupleOperations that could be lowered to ValueTuple construction and destruction, or lowered to assignments to local variables -- but there is no way to know for sure. However, the basic (and relatively common) case of (a, b) = (b, a) should be supported. This requires us to model a Tuple value in dataflow. In the current implementation, we can model it in Deconstruction evaluation, but it's not a first-class concept in the dataflow visitor. This PR stack makes the concept of a Tuple of TValue a first-class concept.

ITupleOperations are treated as Top in the lattice for all cases except for when it is the outer operation on a DeconstructionOperation (i.e. an expression in the form of (a, b) = (b, a)). If the TupleOperation appears elsewhere in the deconstruction (for example, within a ternary that is deconstructed (a, b) = c ? (x, y) : (b, a)), we make no assumption about whether or not a ValueTuple is created.

This PR preserves recursively evaluated direct tuple deconstruction sources before target assignments. Ordinary tuple flow still evaluates elements for effects without guessing whether syntax lowers to ValueTuple fields.

Validation

• Analyzer DataFlow: 85 passed, 2 skipped.
• ILLink DataFlow0: 42 passed; DataFlow1: 41 passed.
• NativeAOT shared ConstructedTypesDataFlow and DeconstructFieldTarget: 2 passed.

Note

This PR was generated with GitHub Copilot assistance.

---

_{Stack created with GitHub Stacks CLI • Give Feedback 💬}

[azure-pipelines]

Azure Pipelines:
Successfully started running 3 pipeline(s).
13 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[dotnet-policy-service]

Tagging subscribers to this area: @agocke, @dotnet/illink
See info in area-owners.md if you want to be subscribed.