[PATCH] phy: ti: twl4030-usb: unwind late probe failures

[Pengpeng Hou]

Hi Mani,

The pm_runtime_put_noidle() path is wrong here: it drops the usage count
without running the suspend callback that powers down the PHY and
disables the regulators.

A replacement also needs to quiesce the IRQ and delayed work before
teardown. The IRQ path can hold an additional runtime-PM reference while
a cable is present, and the PHY is already publicly registered by this
point. Replacing only put_noidle() would not close those lifetime paths.

I'll keep the current patch on hold while reworking the publication and
cleanup ordering rather than resend the one-call substitution.

Thanks,
Pengpeng