Test coverage gap: run_command logic in the recent parity sweep is untested

[jacalata]

Several PRs in the current Classic-parity sweep add argparse-layer tests
but no coverage of the load-bearing logic in run_command where the
parsed args actually get applied to TSC items or normalized. This is a
common pattern across the sweep, not a single-PR issue, so recording
here as a follow-up rather than blocking those PRs.

Concrete gaps

#451 delete syntax normalization. run_command transforms
isinstance(args.workbook, str) -> args.name = args.workbook; args.workbook = True (and same for datasource). This is what makes both
delete Name --workbook and Classic delete --workbook Name produce
equivalent behavior. No test covers the transform; both syntaxes are
verified at the parser layer only. If someone refactors run_command
and drops the normalization, every parser test still passes but Classic
scripts break silently.

#452 editsite attribute application. run_command reads ten new
argparse fields onto SiteItem attributes with three separate patterns:
_str_to_bool coercion for the string-value flags, direct assignment
for the paired boolean flags, and inversion for disable_subscriptions
(from allow_subscriptions). The paired custom_subscription_email +
custom_subscription_email_enabled writes on --subscription-email
(where empty string flips _enabled off) also live in run_command.
None of this is exercised by a mocked server.sites.update call. A
future refactor could drop any of these assignments and no test would
catch it.

#447 creategroup --role. run_command sets
GroupItem.minimum_site_role = args.role before
server.groups.create. Not tested; someone could delete the assignment
and every parser test still passes.

#448 publish --description. run_command sets
new_item.description = args.description on the WorkbookItem or
DatasourceItem before publish. Not tested; same failure mode.

#444 --idp-configuration-id. run_command sets
user_obj.idp_configuration_id = args.idp_configuration_id on every
UserItem in the batch, and apply_cli_overrides clears any pre-existing
auth_setting when the IDP flag is passed. Only the
apply_cli_overrides helper is tested; nothing verifies the attribute
actually reaches the wire.

Suggested shape

Adopt a pattern like tests/commands/test_create_site_users.py uses:
mock the Session/server, mock server.<endpoint>.<method>, invoke
Command.run_command(args), assert the item passed to
server.<endpoint>.<method> has the expected attributes set. Not
required per-flag -- one test per PR covering the main happy path is
enough to lock in the wire-level behavior.

Not tracking here

• Fix #433: accept Classic delete --workbook Name syntax #451 bare --workbook with no value: parser accepts, run_command
  errors out. Already covered at the parser layer; run_command's error
  message is standard.
• Command-level auth and session setup: shared across all commands and
  tested elsewhere.
• Fix #35: --nowait / --silent-progress on createsiteusers #455 createsiteusers: already has run_command coverage
  (test_create_site_users.py); the gap is elsewhere.