parametrizing a fixture with indirect=True does not handle dicts correctly

When using @pytest.mark.parametrize(..., indirect=True) a dict param is handled like a list of its keys, instead of a single param - I have to explicitly wrap it in a list or single-element tuple. Besides making the code less pleasant to read, it also kind of breaks the ruff PT007 linting rule (I set it to always prefer tuples for parametrization), because now I have to use a single-element tuple (which looks even worse than the list)...

Maybe a dict should always be handled as a single element?

import pytest


@pytest.fixture
def fixture_with_param(request):
    print(f'param: {request.param}')
    return request.param['foo'].upper()


@pytest.mark.parametrize('fixture_with_param', {'foo': 'bar', 'hello': 'world'}, indirect=True)
def test_bad_param(fixture_with_param):
    assert fixture_with_param == 'BAR'


@pytest.mark.parametrize('fixture_with_param', [{'foo': 'bar', 'hello': 'world'}], indirect=True)
def test_good_param(fixture_with_param):
    assert fixture_with_param == 'BAR'

$ pytest
================================================= test session starts ==================================================
platform linux -- Python 3.11.6, pytest-7.4.3, pluggy-1.3.0
rootdir: /tmp/empty
collected 3 items

test_indirect_param.py EE.                                                                                       [100%]

======================================================== ERRORS ========================================================
________________________________________ ERROR at setup of test_bad_param[foo] _________________________________________

request = <SubRequest 'fixture_with_param' for <Function test_bad_param[foo]>>

    @pytest.fixture
    def fixture_with_param(request):
        print(f'param: {request.param} ({type(request.param)})')
>       return request.param['foo'].upper()
E       TypeError: string indices must be integers, not 'str'

test_indirect_param.py:7: TypeError
------------------------------------------------ Captured stdout setup -------------------------------------------------
param: foo (<class 'str'>)
_______________________________________ ERROR at setup of test_bad_param[hello] ________________________________________

request = <SubRequest 'fixture_with_param' for <Function test_bad_param[hello]>>

    @pytest.fixture
    def fixture_with_param(request):
        print(f'param: {request.param} ({type(request.param)})')
>       return request.param['foo'].upper()
E       TypeError: string indices must be integers, not 'str'

test_indirect_param.py:7: TypeError
------------------------------------------------ Captured stdout setup -------------------------------------------------
param: hello (<class 'str'>)
=============================================== short test summary info ================================================
ERROR test_indirect_param.py::test_bad_param[foo] - TypeError: string indices must be integers, not 'str'
ERROR test_indirect_param.py::test_bad_param[hello] - TypeError: string indices must be integers, not 'str'
============================================= 1 passed, 2 errors in 0.02s ==============================================

$ python --version
Python 3.11.6

$ pip list
Package    Version
---------- -------
iniconfig  2.0.0
packaging  23.2
pip        23.3.1
pluggy     1.3.0
pytest     7.4.3
setuptools 65.5.0

Start by reproducing the indirect=True examples from the issue with pytest 7.4.3, then trace how parametrization converts the dict into individual parameters. Determine the intended handling for dict values and add regression coverage; done means the dict reaches the fixture as one parameter while existing parametrization behavior remains intact.