[Upstream] Cannot switch to Laser-based tensors

[mratsim]

Laser-based tensors use either a raw pointer to be able to plug in buffers coming from external libraries (Numpy, PyTorch, Tensorflow) and also exchange buffer with zero-copy. Or they use a seq for Nim ref types or strings.

Arraymancer/src/laser/tensor/datatypes.nim

Lines 12 to 30 in d3f078a

	type
	RawImmutableView*[T] = distinct ptr UncheckedArray[T]
	RawMutableView*[T] = distinct ptr UncheckedArray[T]
	Metadata* = DynamicStackArray[int]
	Tensor*[T] = object # Total stack: 128 bytes = 2 cache-lines
	shape*: Metadata # 56 bytes
	strides*: Metadata # 56 bytes
	offset*: int # 8 bytes
	storage*: CpuStorage[T] # 8 bytes
	CpuStorage*{.shallow.}[T] = ref object # Total heap: 25 bytes = 1 cache-line
	when supportsCopyMem(T):
	raw_buffer*: ptr UncheckedArray[T] # 8 bytes
	memalloc*: pointer # 8 bytes
	isMemOwner*: bool # 1 byte
	else: # Tensors of strings, other ref types or non-trivial destructors
	raw_buffer*: seq[T] # 8 bytes (16 for seq v2 backed by destructors?)

Unfortunately supportsCopyMem doesn't work properly in type section, making Laser Dead On Arrival :/. nim-lang/Nim#13095

[mratsim]

Original issue fixed but now there is a new one it's not a showstopper but always having to put the generic type is annoying and even if Arraymancer is clean, users proc can still trigger it: nim-lang/Nim#13193