Get matplotlib working with upstream Python

This patch adds support for using matplotlib plus its Gtk backend with
`--config=k8_upstream_python`. Unfortunately, the Tk backend doesn't
work because of how Python is packaged. See the following issue for
more information:
https://github.com/matplotlib/matplotlib/issues/23074

Separately, we need to patch pygobject and matplotlib for various
hermeticity reasons. To accomplish this, I've added patching support
for wheels downloaded via rules_python. I should have set it up to use
annotations (the same mechanism we use for injecting deps), but it was
a little cumbersome. Annotations are not set up for using in arguments
to repository rules. I instead opted for a separate JSON file. You can
find it at `tools/python/patches.json`.

You can try the two new example programs:

    $ bazel run --config=k8_upstream_python //build_tests:matplotlib_example
    $ bazel run --config=k8_upstream_python //build_tests:pygobject_example

Signed-off-by: Philipp Schrader <philipp.schrader@gmail.com>
Change-Id: I3c4e8648a8012aa23dedd53610e82d590d1c409d
diff --git a/build_tests/BUILD b/build_tests/BUILD
index 7c88992..718e817 100644
--- a/build_tests/BUILD
+++ b/build_tests/BUILD
@@ -232,3 +232,20 @@
         "@pip//scipy",
     ],
 )
+
+py_binary(
+    name = "pygobject_example",
+    srcs = ["pygobject_example.py"],
+    deps = [
+        "@pip//pygobject",
+    ],
+)
+
+py_binary(
+    name = "matplotlib_example",
+    srcs = ["matplotlib_example.py"],
+    deps = [
+        "@pip//matplotlib",
+        "@pip//pygobject",
+    ],
+)