Add contextlib support to the pyhf.schema API

[kratsg]

instead of having users overwriting the value by assignment, do you think there is value in offering a pyhf.utils.update_schema_path or something that performs this operation through the API?

Thanks @kratsg for taking up my suggestion!

I second @matthewfeickert's suggestion to change schemas via function call rather than assignment. It is much simpler to add functionality to a function if it ever becomes necessary, than to replace an entire existing module with a class instance with a property.

I'd even go so far to say that a context manager doubling as an update function would be ideal IMO:

# in pyhf.utils
_SCHEMAS = Path(...)
class use_schema_path: # snake_case to remind of function-like usage
    def  __init__(self, path):
        global _SCHEMAS
        self._old_schemas = _SCHEMAS
        _SCHEMAS = pathlib.Path(path)
    def __enter__(self):
        pass
    def __exit__(self, *args, **kwargs):
        global _SCHEMAS
        _SCHEMAS = self._old_schemas

which can still be called as a function (only executing __init__), so short scripts etc. are not forced to use with blocks.
But it can also be used like so:

def make_my_workspace(spec):
    with pyhf.utils.use_schema_path('/my/very/special/schemas'):
        return pyhf.Workspace(spec)

So as a user writing code on top of pyhf, I don't have to worry about resesetting the global variable, the CM does it for me, and there are fewer mistakes to make.

Originally posted by @lhenkelm in #1753 (comment)

[kratsg]

This is on top of #1609 .