test(backup): Create backup version snapshot tests (#58173)

The goal for the export.json feature is to support two minor versions
(since we're using CalVer, roughly equivalent to two months) of releases
when importing. That is to say, if the latest release is 23.10.0, and
the previous two releases were 23.9.0 and 23.8.0, then a self-hosted
release running code from [email protected] or self-hosted@latest (aka
nightly) should be able to successfully import JSON backup data
generated by [email protected] and [email protected], but not
necessarily [email protected].

To support this, we add a new file `test_releases.py` that provides a
snapshot test of an "exhaustive" (aka, has at least one of every
exportable model) export. Every time we cut a new release, we will add a
new commit directly after the release commit that renames the
`test_at_head` snapshot to `test_at_<RELEASED_VERSION>`, and creates a
test case for that version (see `test_at_23_9_1` in this PR for one such
example). Test cases older than the 2 version support window will
simultaneously be removed.

In the future, we plan to automate this "release test rotation" process
via GitHub actions, but for now the ospo team will take care to do this
manually for every release.

Some other notes of interest:
- Exports cannot be compared for equality using simple string
comparisons, since they have special comparators for cases like
passwords, dates that may or may not have been updated, etc (see
comparators.py for greater detail). To accommodate this, we extend the
`insta_snapshot` fixture with an extra argument,
`inequality_comparator`, that allows a custom comparison lambda to be
passed in for such cases.
- The `test_at_23_9_1` snapshot was generated by taking an early version
of this PR, cherry-picking it onto the `23.9.1` release tag, generating
a new snapshot there, then copying the result back into the main commit.

Closes getsentry/team-ospo#197

src/sentry/testutils/pytest/fixtures.py

@@ -11,7 +11,8 @@
 import sys
 from concurrent.futures import ThreadPoolExecutor
 from datetime import datetime
-from typing import Optional
+from string import Template
+from typing import Optional, Tuple
 
 import pytest
 import requests
@@ -331,9 +332,24 @@ def ignore_aliases(self, data):
         return True
 
 
+def read_snapshot_file(reference_file: str) -> Tuple[str, str]:
+    with open(reference_file, encoding="utf-8") as f:
+        match = _yaml_snap_re.match(f.read())
+        if match is None:
+            raise OSError()
+
+        header, refval = match.groups()
+        return (header, refval)
+
+
 @pytest.fixture
 def insta_snapshot(request, log):
-    def inner(output, reference_file=None, subname=None):
+    def inner(
+        output,
+        reference_file=None,
+        subname=None,
+        inequality_comparator=lambda refval, output: refval != output,
+    ):
         if reference_file is None:
             name = request.node.name
             for c in UNSAFE_PATH_CHARS:
@@ -362,18 +378,15 @@ def inner(output, reference_file=None, subname=None):
             )
 
         try:
-            with open(reference_file, encoding="utf-8") as f:
-                match = _yaml_snap_re.match(f.read())
-                if match is None:
-                    raise OSError()
-                _header, refval = match.groups()
+            _, refval = read_snapshot_file(reference_file)
         except OSError:
             refval = ""
 
         refval = refval.rstrip()
         output = output.rstrip()
+        is_unequal = inequality_comparator(refval, output)
 
-        if _snapshot_writeback is not None and refval != output:
+        if _snapshot_writeback is not None and is_unequal:
             if not os.path.isdir(os.path.dirname(reference_file)):
                 os.makedirs(os.path.dirname(reference_file))
             source = os.path.realpath(str(request.node.fspath))
@@ -397,27 +410,47 @@ def inner(output, reference_file=None, subname=None):
                         output,
                     )
                 )
-        elif refval != output:
+        elif is_unequal:
             __tracebackhide__ = True
-            _print_insta_diff(reference_file, refval, output)
+            if isinstance(is_unequal, str):
+                _print_custom_insta_diff(reference_file, is_unequal)
+            else:
+                _print_insta_diff(reference_file, refval, output)
 
     yield inner
 
 
+INSTA_DIFF_TEMPLATE = Template(
+    """~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+Snapshot $reference_file changed!
+
+
+Re-run pytest with SENTRY_SNAPSHOTS_WRITEBACK=new and then use 'make review-python-snapshots' to review.
+
+Or: Use SENTRY_SNAPSHOTS_WRITEBACK=1 to update snapshots directly.
+
+
+$diff_text
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+"""
+)
+
+
 def _print_insta_diff(reference_file, a, b):
     __tracebackhide__ = True
     pytest.fail(
-        "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n"
-        "Snapshot {} changed!\n\n"
-        "Re-run pytest with SENTRY_SNAPSHOTS_WRITEBACK=new and then use 'make review-python-snapshots' to review.\n"
-        "Or: Use SENTRY_SNAPSHOTS_WRITEBACK=1 to update snapshots directly.\n\n"
-        "{}\n"
-        "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n".format(
-            reference_file, "\n".join(difflib.unified_diff(a.splitlines(), b.splitlines()))
+        INSTA_DIFF_TEMPLATE.substitute(
+            reference_file=reference_file,
+            diff_text="\n".join(difflib.unified_diff(a.splitlines(), b.splitlines())),
         )
     )
 
 
+def _print_custom_insta_diff(reference_file, diff_text):
+    __tracebackhide__ = True
+    pytest.fail(INSTA_DIFF_TEMPLATE.substitute(reference_file=reference_file, diff_text=diff_text))
+
+
 @pytest.fixture
 def call_snuba(settings):
     def inner(endpoint):