Change dive of match clauses to use the outer range, including the match head pattern

[baronfel]

When working on a new codefix I discovered that syntax tree walks of match clauses don't seem to be triggering calls to VisitMatchClause. This seems to be because the range passed into the dive is the range of the guard expression + right hand side of the clause, when I believe it should be the overall range of the match clause. Because of this inconsistency, I think no range matches the range checks in pick, and so the VisitMatchClause member is never called.

[kerams]

Yup, came to the same conclusion in #12930

[baronfel]

Aha! Thank you both for the inspiration. I thought I saw something similar in FsAutoComplete around the walker not walking some nodes, and I thought I was going batty.

[kerams]

Ping, can we get this in? I reckon this will open the door to some improvements to completions in match clauses.

[baronfel]

As an additional sanity check, I looked at the implementation of traverseSynMatchClause, which we would start calling more often if this PR is merged. That function does in fact dive into every part of the match clause: pattern, guard and right-hand-side, and so it is already expecting to be called in this way.