The previous implementation was using the following heuristic
```lean
-- heuristic: use non-dependent arrows only if possible for whole group to avoid
-- noisy mix like `(α : Type) → Type → (γ : Type) → ...`.
let dependent := curNames.any fun n => hasIdent n.getId stxBody
```
The result produced by this heuristic was **not** producing an
accidental name capture, but I agree
it was confusing to have `∀ (a : True), ∃ a, a = a : Prop` instead of
`True → ∃ a, a = a : Prop` since there is no dependency.
AFAICT, all examples affected by this commit have a better output now.
cc @digma0 @kha
Also make it dependent on presence of `missing` instead of parse error,
which means that messages from complete commands that are immediately followed
by parse errors are not filtered out anymore
When `pp.safe_shadowing` is set to true, we still use the
suggested name if the "body" does not contain a free variable with the
suggested name. This is the approach used in Lean 3, and I think it
improved the result in all affected tests.
The implementation was simple. The only nasty case was `delabAppMatch`.
The main motivation for this feature was hovering information such as
```lean
f : {α_1 : Type} → α_1 → α_1
```
when hovering over the `f` at
```lean
def g (α : Type) (a : α) :=
f a
```
With `safe_shadowing`, we get the nicer
```lean
f : {α : Type} → α → α
```
cc @Kha
Now, the following example produces a syntax error.
```lean
macro "foo!" x:term : term => `($x + 1)
check id foo! 10
```
@Kha, I think the heuristic is simple and defensible.
If the new syntax starts and ends with token, than the precedence is
`maxPrec`. Otherwise, it is `leadPrec`.
see #180
@Kha Before this commit, we were producing the error "expected command" at the `let` token
```lean
check id let x := 1; x
```
The new error is "expected command, but found term; this error may be
due to parsing precedence levels, consider parenthesizing the term".
The example above looks artificial, but it will happen all the time as
users start to define their own notation.
