[wip] Possible reversion of 3283 (Tuple<...> regression with Item* and Rest properties)

[dsyme]

It may be the best thing is to revert #3283 because of the gradual emergence of more serious cases of regressions in #3729

We should probably have a final think about this before doing it. The more expensive option mentioned in #3729 is possible, but takes significant dev and test time.

In particular his change in itself will break code that was written in the interim, including:

1. code changed to account for Breaking change in tuples after VS 15.4 update #3729
2. code using Tuple.Create and explicitly annotating with a return type of
3. new code assuming a total equivalence between F#-authored declarations of Tuple<A,B> and A*B.

I've marked this as WIP since a reversion that breaks code makes me feel queasy.

[cartermp]

I agree that this is a rather awkward position to be in, as it is unknown how much code has either:

• Taken advantage of the new behavior as new code (not a workaround)
• Used the new behavior to work around Breaking change in tuples after VS 15.4 update #3729

But assuming that we do go this route, we should also have a quick test for the regression, which could then be updated when we complete the work to make them the same and allow for accessing those properties.

[dsyme]

I'm closing this in favour of forward-moving resolution #4034. We can reopen if we decide to move backwards instead