Adding memory provider test fakes

Summary:
Part of
* https://github.com/meta-llama/llama-stack/issues/436

This change adds a minimalistic support to creating memory provider test fake. For more details about the approach, follow the issue link from above.

Test Plan:
Run tests using the "test_fake" mark:
```
pytest llama_stack/providers/tests/memory/test_memory.py -m "test_fake"
/opt/homebrew/Caskroom/miniconda/base/envs/llama-stack/lib/python3.11/site-packages/pytest_asyncio/plugin.py:208: PytestDeprecationWarning: The configuration option "asyncio_default_fixture_loop_scope" is unset.
The event loop scope for asynchronous fixtures will default to the fixture caching scope. Future versions of pytest-asyncio will default the loop scope for asynchronous fixtures to function scope. Set the default fixture loop scope explicitly in order to avoid unexpected behavior in the future. Valid fixture loop scopes are: "function", "class", "module", "package", "session"

  warnings.warn(PytestDeprecationWarning(_DEFAULT_FIXTURE_LOOP_SCOPE_UNSET))
====================================================================================================== test session starts ======================================================================================================
platform darwin -- Python 3.11.10, pytest-8.3.3, pluggy-1.5.0
rootdir: /llama-stack
configfile: pyproject.toml
plugins: asyncio-0.24.0, anyio-4.6.2.post1
asyncio: mode=Mode.STRICT, default_loop_scope=None
collected 18 items / 15 deselected / 3 selected

llama_stack/providers/tests/memory/test_memory.py ...                                                                                                                                                                     [100%]

========================================================================================= 3 passed, 15 deselected, 6 warnings in 0.03s ==========================================================================================
```

llama_stack/providers/tests/resolver.py

@@ -14,7 +14,10 @@ from llama_stack.distribution.build import print_pip_install_help
 from llama_stack.distribution.configure import parse_and_maybe_upgrade_config
 from llama_stack.distribution.distribution import get_provider_registry
 from llama_stack.distribution.request_headers import set_request_provider_data
-from llama_stack.distribution.resolver import resolve_remote_stack_impls
+from llama_stack.distribution.resolver import (
+    resolve_remote_stack_impls,
+    resolve_test_fake_stack_impls,
+)
 from llama_stack.distribution.stack import construct_stack
 from llama_stack.providers.utils.kvstore import SqliteKVStoreConfig
 
@@ -51,16 +54,26 @@ async def construct_stack_for_test(
     )
     run_config = parse_and_maybe_upgrade_config(run_config)
     try:
-        remote_config = remote_provider_config(run_config)
-        if not remote_config:
-            # TODO: add to provider registry by creating interesting mocks or fakes
-            impls = await construct_stack(run_config, get_provider_registry())
-        else:
-            # we don't register resources for a remote stack as part of the fixture setup
-            # because the stack is already "up". if a test needs to register resources, it
-            # can do so manually always.
+        impls = None
 
-            impls = await resolve_remote_stack_impls(remote_config, run_config.apis)
+        # "Resolve" implementations when using test::test-fake providers.
+        # This is actually injecting test fakes as resolved API implementations.
+        test_fake_config = test_fake_provider_config(run_config)
+        if test_fake_config:
+            impls = await resolve_test_fake_stack_impls(test_fake_config)
+
+        # Resolve implementations when using test::remove providers
+        if not impls:
+            remote_config = remote_provider_config(run_config)
+            if remote_config:
+                # we don't register resources for a remote stack as part of the fixture setup
+                # because the stack is already "up". if a test needs to register resources, it
+                # can do so manually always.
+                impls = await resolve_remote_stack_impls(remote_config, run_config.apis)
+
+        # In case none of the above happened, resolve implementations as normal providers
+        if not impls:
+            impls = await construct_stack(run_config, get_provider_registry())
 
         test_stack = TestStack(impls=impls, run_config=run_config)
     except ModuleNotFoundError as e:
@@ -75,6 +88,28 @@ async def construct_stack_for_test(
     return test_stack
 
 
+# In case when we want to run multiple test fakes, then we need to
+# make sure the stack contains only test fake providers as after this
+# we will be calling resolve_test_fake_stack_impls() which cannot resolve
+# any other provider type. This could be refactored in case we need to change this.
+def test_fake_provider_config(
+    run_config: StackRunConfig,
+) -> Optional[RemoteProviderConfig]:
+    test_fake_config = None
+    has_non_test_fake = False
+    for api_providers in run_config.providers.values():
+        for provider in api_providers:
+            if provider.provider_type == "test::test-fake":
+                test_fake_config = TestFakeProviderConfig(**provider.config)
+            else:
+                has_non_test_fake = True
+
+    if test_fake_config:
+        assert not has_non_test_fake, "Test fake stack cannot have non-fake providers"
+
+    return test_fake_config
+
+
 def remote_provider_config(
     run_config: StackRunConfig,
 ) -> Optional[RemoteProviderConfig]: