usb: dwc2: remove WARN in dwc2_hcd_save_data_toggle

usb: dwc2: remove WARN in dwc2_hcd_save_data_toggle

The WARN() in dwc2_hcd_save_data_toggle() was introduced in
commit 62943b7dfa35 ("usb: dwc2: host: fix the data toggle error in
full speed descriptor dma"), it looks like the WARN() is to ensure
proper usage of dwc2_hcd_save_data_toggle(): either qtd is provided
for control eps or qh is provided for non-control eps. This check is
good even if there's no such improper usage in current code. But the
WARN() usage in driver is discouraged nowadays: imagine there is an
improper usage, then kernel panic due to warn if 'panic_on_warn' is
enabled.

While emitting the err msg for improper usage is still valueable, so
let's replace the WARN with check and dev_err().

At the same time, it looks a bit strange we check !chan after
dereference of this pointer with
"if (chan->ep_type != USB_ENDPOINT_XFER_CONTROL)".

In fact, when entering the dwc2_hcd_save_data_toggle(), the chan won't
be NULL, because its caller or indirect caller has ensured this,
specifically, it's checked with below line in dwc2_hc_n_intr()

if (!chan) {
    dev_err(hsotg->dev, "## hc_ptr_array for channel is NULL ##\n");
    return;
}

This addresses the following issue reported by klocwork tool:
  - Suspicious dereference of pointer 'chan' before NULL check at
    line 518

Suggested-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Link: https://patch.msgid.link/20260520133711.14410-1-jszhang@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>