Support recording kernel launches of Launch.launch() onto wp.Tape()

[shi-eric]

Currently, Launch.launch() only launches the kernel. We need to add in the parts from wp.launch() that add the kernel to the tape and does gradient debugging checks:

warp/warp/context.py

Lines 5594 to 5605 in 56a290b

	# record on tape if one is active
	if runtime.tape and record_tape:
	# record file, lineno, func as metadata
	frame = inspect.currentframe().f_back
	caller = {"file": frame.f_code.co_filename, "lineno": frame.f_lineno, "func": frame.f_code.co_name}
	runtime.tape.record_launch(
	kernel, dim, max_blocks, inputs, outputs, device, block_dim, metadata={"caller": caller}
	)
	# detect illegal inter-kernel read/write access patterns if verification flag is set
	if warp.config.verify_autograd_array_access:
	runtime.tape._check_kernel_array_access(kernel, fwd_args)