Treat the CSV as pure YAML

We used to treat the CSV as an unknown text format,
but this has grown cumbersome and is violated later
in the bundle building process where the file is
read / rewrite by yaml.dump. Changing to do the same
thing here.

doozerlib/distgit.py

@@ -2243,45 +2243,22 @@ def _update_csv(self, version, release):
             # We need to inject "skips" versions for https://issues.redhat.com/browse/OCPBUGS-6066 .
             # We have the versions, but it needs to be written into the CSV under spec.skips.
             # First, find the "name" of this plugin that precedes the version. If everything
-            # is correctly replaced in the csv by art-config.yml, then metadata.name - spec.version
+            # is correctly replaced in the CSV by art-config.yml, then metadata.name - spec.version
             # should leave us with the name the operator uses.
             csv_obj = yaml.safe_load(pathlib.Path(csv_file).read_text())
             olm_name = csv_obj['metadata']['name']  # "nfd.4.11.0-202205301910"
             olm_version = csv_obj['spec']['version']  # "4.11.0-202205301910"
+
+            if not olm_name.endswith(olm_version):
+                raise IOError(f'Expected {self.name} CSV metadata.name field ("{olm_name}" after rebase) to be suffixed by spec.version ("{olm_version}" after rebase). art-config.yml / upstream CSV metadata may be incorrect.')
+
             olm_name_prefix = olm_name[:-1 * len(olm_version)]  # "nfd."
 
             # Inject the skips..
-            # If we re-wrote this as YAML using the yaml package, it would be easy,
-            # but historically we haven't. There's probably a reason for this, so we'll just be
-            # injecting this using awkward text operations. The only real challenge is finding
-            # the indentation being used in the YAML so we can inject with consistent indentation.
-            # To find that indentation, search for "version: " stanza. There will be multiple,
-            # but the one with the least indentation will be the one under spec: .
-            csv_lines = pathlib.Path(csv_file).read_text().splitlines(keepends=False)
-            version_regex = re.compile(r'(\s+)"?version:.*')
-            indentation_step = None
-            version_line_index = -1
-            for line_index, line in enumerate(csv_lines):
-                match = version_regex.match(line)
-                if match:
-                    potential_indentation = match.group(1)
-                    if indentation_step is None:
-                        indentation_step = potential_indentation  # Isolate the indentation whitespace
-                        version_line_index = line_index
-                    elif len(potential_indentation) < len(indentation_step) and len(potential_indentation) > 0:
-                        indentation_step = potential_indentation
-                        version_line_index = line_index
-
-            if version_line_index < 0:
-                raise IOError(f'Unable to find version: stanza in CSV file: {self.name} / {csv_file}')
-
-            new_csv_lines = list(csv_lines)
-            new_csv_lines.insert(version_line_index + 1, f'{indentation_step}skips:')  # inject skips after "version: " line.
-            for old_version in previous_build_versions:
-                # Add a list of versions after 'skips'.
-                new_csv_lines.insert(version_line_index + 2, f'{indentation_step}- {olm_name_prefix}{old_version}')
-
-            pathlib.Path(csv_file).write_text('\n'.join(new_csv_lines))
+            csv_obj['spec']['skips'] = [f'{olm_name_prefix}{old_version}' for old_version in previous_build_versions]
+
+            # Re-write the CSV content.
+            pathlib.Path(csv_file).write_text(yaml.dump(csv_obj))
 
     def _update_environment_variables(self, update_envs, filename='Dockerfile'):
         """