pwm changes for 6.8, take 2

The first commit fixes a duplicate cleanup in an error path introduced
in pwm/for-6.8-rc1~13.

The second cares for an out-of-bounds access. In practise it doesn't
happen---otherwise someone would have noticed since v5.17-rc1 I
guess---because the device tree binding for the two drivers using
of_pwm_single_xlate() only have args->args_count == 1. A device-tree
that doesn't conform to the respective bindings could trigger that
easily however.

The third and last one corrects the request callback of the jz4740 pwm
driver which used dev_err_probe() long after .probe() completed. This is
conceptually wrong because dev_err_probe() might call
device_set_deferred_probe_reason() which is nonsensical after the driver
is bound.
-----BEGIN PGP SIGNATURE-----

iQEzBAABCgAdFiEEP4GsaTp6HlmJrf7Tj4D7WH0S/k4FAmWo3dwACgkQj4D7WH0S
/k5L3Qf8Db1uwUEGFXjjMIS82djW6u8rAEu1bEVf0yqhFKB3zo/PO4b1bfTHvKGo
SwSflvZ2UB+bcby1yIZTP1yUbDH1O3WFfxgYwdUR715us/iMCV5sYvuNagi0Fp8O
edR1Ntr/CmaU5VeDyml7ZUIHj7oZ50hd1dvvEugvKQKEVvW8yHBR2OEuqI38cIyK
ItsQPrN9ld7PXzE0UFWMJpi3lbo+Q4KEYbx0yzBmBGogLXi+nRCIGtMB5KzdUSvf
K4mQPHDmUyHsviXyjXcywst7sL1ebUcI3FicvCf4by/Sek5jZQGtLoyRS0m4FkRw
V/TQMiLhU5iKr84wPQUvw6ZYmcXxog==
=BP0a
-----END PGP SIGNATURE-----