tests/xe/debugfs: fix double-free in test_base

Looks to incorrectly free the xe_dev->config, even though there is still
a cached entry pointing to it. The test fixture will also call
xe_device_put() at the end, freeing the same pointer (might now point to
valid memory), which hopefully explains why we see strange looking
crashes in CI:

Starting subtest: base
Subtest base: SUCCESS (0.005s)
Received signal SIGABRT.
Stack trace:
 #0 [fatal_sig_handler+0x10f]
 #1 [killpg+0x40]
 #2 [gsignal+0xcb]
 #3 [abort+0x12b]
 #4 [__fsetlocking+0x42e]
 #5 [pthread_attr_setschedparam+0x54c]
 #6 [pthread_attr_setschedparam+0xd28]
 #7 [pthread_attr_setschedparam+0x2ed3]
 #8 [__libc_malloc+0x74]
 #9 [_IO_file_doallocate+0x94]
 #10 [_IO_doallocbuf+0x50]
 #11 [_IO_file_overflow+0x1b0]
 #12 [_IO_file_xsputn+0xe5]
 #13 [psiginfo+0x13512]
 #14 [igt_kmsg+0xc9]
 #15 [igt_exit+0xd3]
 #16 [main+0x44]
 #17 [__libc_start_main+0xf3]
 #18 [_start+0x2e]

Keep the xe_device stuff in test_base, since that looks be the only user
here.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Zbigniew Kempczynski <zbigniew.kempczynski@intel.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>