Still getting a grip with how mypyc does things, but I feel like I am at a spot
where I can actually start doing things now.

Since the AsPattern binds a value to name, it is very coupled to the pattern
that is actually being built. There is (in general) 3 stages to a typical
pattern match:

1. Converting the expression to a value/register
2. Checking if the match was successful
3. Assigning the value to a variable if it is an AsPattern

For this to work though, steps 2 and 3 need to be done at the same time,
because the jump must jump to a block which sets the variable.
Maybe I am wrong, and I can just assign to a variable, and if the match
did not pass, it will simply be reassigned.

For some reason PyMapping_Check returns true for string types, dont know why.
This probably means I have to to do a subclass check.

We need to do a check to see that the list is not too short for the number of
patterns we have. Currently there is a failing test, which will be fixed momentarily.

Turns out that the `-ffat-lto-objects` flag is only available in GCC, and clang
will complain if there are any unused optimization flags passed. Adding the
`-Wno-ignored-optimization-argument` flag seems to fix this.

These lines will be moved to a separate PR so that they can be reviewed
separately from the `match` stuff going on here.

Since the `Py_TPFLAGS_MAPPING` and `Py_TPFLAGS_SEQUENCE` are specific to the
new pattern matching feature, they are not defined for Python 3.9 and below.
Now they are defined if they don't exist, which should fix the issue.

It would appear that 9125155 is now causing issues with the IR tests.
Reverting that commit seems to fix it, but I cannot seem to remember why I
added it in the first place, and the build artifacts have long since expired.

* Use faster "isinstance" op if the expression is a built-in type

* Add type annotations to run tests

* Add type annotated IR example

* Add todo comments