usb: dwc3-msm: Fix memory leak of dwc3 child device

In case of failed probe or driver removal, the dwc3 child device
is left with an extra reference count and hence does not get
cleaned up properly, resulting in a memory leak. kmemleak reports
the following (and others):

unreferenced object 0xffffffdf5934a600 (size 128):
  comm "kworker/2:1", pid 83, jiffies 4294938474 (age 2208.110s)
  hex dump (first 32 bytes):
    00 00 60 0a 00 00 00 00 ff cc 60 0a 00 00 00 00  ..`.......`.....
    a0 5f 2b 7d df ff ff ff 00 02 00 00 00 00 00 00  ._+}............
  backtrace:
    [<ffffff9aeb4a0cb8>] kmemleak_alloc+0x60/0xc0
    [<ffffff9aea63b51c>] __kmalloc+0x220/0x3bc
    [<ffffff9aeafdd5b8>] of_device_alloc+0x98/0x1a8
    [<ffffff9aeafdd87c>] of_platform_device_create_pdata+0x58/0xcc
    [<ffffff9aeafddc38>] of_platform_bus_create+0x220/0x438
    [<ffffff9aeafdded8>] of_platform_populate+0x88/0x104
    [<ffffff9aeacf10f8>] dwc3_msm_probe+0xc7c/0x12f0
    [<ffffff9aeab46514>] platform_drv_probe+0x5c/0xb0
    [<ffffff9aeab43994>] driver_probe_device+0x2e8/0x41c
    [<ffffff9aeab4440c>] __device_attach_driver+0xcc/0x144
    [<ffffff9aeab41b00>] bus_for_each_drv+0x80/0xc8
    [<ffffff9aeab43bf0>] __device_attach+0xa4/0x154
    [<ffffff9aeab43cc0>] device_initial_probe+0x20/0x2c
    [<ffffff9aeab41ca4>] bus_probe_device+0x34/0x8c
    [<ffffff9aeab441a8>] deferred_probe_work_func+0xcc/0x16c
    [<ffffff9aea4d1fe4>] process_one_work+0x1c0/0x3d4

Plug this by adding a call to platform_device_put() which balances
the implicit get_device() from the of_find_device_by_node() call
done during probe.

Change-Id: I1e590cfe14ebbb66005b41f6dfa6a56df9489288
Signed-off-by: Jack Pham <jackp@codeaurora.org>