Enable KeyPath/AnyKeyPath/PartialKeyPath/WritableKeyPath in Embedded Swift, but
for compile-time use only:
- Add keypath optimizations into the mandatory optimizations pipeline
- Allow keypath optimizations to look through begin_borrow, to make them work
even in OSSA.
- If a use of a KeyPath doesn't optimize away, diagnose in PerformanceDiagnostics
- Make UnsafePointer.pointer(to:) transparent to allow the keypath optimization
to happen in the callers of UnsafePointer.pointer(to:).
Try to replace a begin_borrow with its owned operand.
This is either possible if the borrowed value (or a forwarded value if it) is copied:
```
%1 = begin_borrow %0
%2 = struct_extract %1 // a chain of forwarding instructions
%3 = copy_value %1
// ... uses of %3
end_borrow %1
```
->
```
%1 = copy_value %0
%3 = destructure_struct %0 // owned version of the forwarding instructions
// ... uses of %3
```
Or if the borrowed value is destroyed immediately after the borrow scope:
```
%1 = begin_borrow %0
%2 = struct_extract %1 // a chain of forwarding instructions
// ... uses of %2
end_borrow %1
destroy_value %1 // the only other use of %0 beside begin_borrow
```
->
```
%2 = destructure_struct %0 // owned version of the forwarding instructions
// ... uses of %2
destroy_value %2
```
For a redundant pair of pointer-address conversions, e.g.
%2 = address_to_pointer %1
%3 = pointer_to_address %2 [strict]
replace all uses of %3 with %1.
Look through `upcast` and `init_existential_ref` instructions and replace the operand of this cast instruction with the original value.
For example:
```
%2 = upcast %1 : $Derived to $Base
%3 = init_existential_ref %2 : $Base : $Base, $AnyObject
checked_cast_br %3 : $AnyObject to Derived, bb1, bb2
```
This makes it more likely that the cast can be constant folded because the source operand's type is more accurate.
In the example above, the cast reduces to
```
checked_cast_br %1 : $Derived to Derived, bb1, bb2
```
which can be trivially folded to always-succeeds.
Found while looking at `_SwiftDeferredNSDictionary.bridgeValues()`
Optimize the sequence
```
%1 = init_enum_data_addr %enum_addr, #someCaseWithPayload
store %payload to %1
inject_enum_addr %enum_addr, #someCaseWithPayload
```
to
```
%1 = enum $E, #someCaseWithPayload, %payload
store %1 to %enum_addr
```
This sequence of three instructions must appear in consecutive order.
But usually this is the case, because it's generated this way by SILGen.
This comes up in the code for constructing an empty string literal.
With this optimization it's possible to statically initialize empty string global variables.
* move the apply of partial_apply transformation from simplify-apply to simplify-partial_apply
* delete dead partial_apply instructions
* devirtualize apply, try_apply and begin_apply
Eliminate the redundant instruction pair
```
%t = tuple (%0, %1, %2)
(%3, %4, %5) = destructure_tuple %t
```
and replace the results %3, %4, %5 with %0, %1, %2, respectively.
The same for structs.