Add ability to clear loaded schema cache.

[WilliamJamieson]

The schemas opened via asdf.schema.load_schema are cached using:

asdf/asdf/schema.py

Lines 462 to 488 in a346a6b

	@lru_cache
	def _load_schema_cached(url, resolver, resolve_references):
	loader = _make_schema_loader(resolver)
	schema, url = loader(url)
	if resolve_references:
	def resolve_refs(node, json_id):
	if json_id is None:
	json_id = url
	if isinstance(node, dict) and "$ref" in node:
	suburl_base, suburl_fragment = _safe_resolve(resolver, json_id, node["$ref"])
	if suburl_base == url or suburl_base == schema.get("id"):
	# This is a local ref, which we'll resolve in both cases.
	subschema = schema
	else:
	subschema = load_schema(suburl_base, resolver, True)
	return reference.resolve_fragment(subschema, suburl_fragment)
	return node
	schema = treeutil.walk_and_modify(schema, resolve_refs)
	return schema

This can present a problem when working with schema resources that are not included in asdf under normal operations. During testing I want to be able to load those schemas, but later I want loading those schemas to fail. This is because otherwise I end up with a situation where the order that tests are run influences whether or not tests pass or fail.

Thus it would be useful to have a mechanism to clear this cache without having to access it through a private API with asdf.schema._load_schema_cached.cache_clear().

Alternately, we may want to make the schema cache tied to the current AsdfConfig so things like what I am trying to do can be done within a special AsdfConfig context.