Some data movement ops don't pass memory config

[odjuricicTT]

E.g. reshape:
https://github.com/tenstorrent/tt-mlir/blob/main/runtime/lib/ttnn/operations/data_movement/reshape.cpp#L16
Should pass the memory config optional parameter when invoked.

In order for optimizer to work, every op that supports it needs to either pass memory_config or the output tensor.

[odjuricicTT]

@jnie-TT Can you take a look at this?

[sdjordjevicTT]

@mtopalovicTT @svuckovicTT @azecevicTT @jserbedzijaTT FYI

This is probably on us @odjuricicTT, most of the ops that don't have memory_config parameter mostly don't have memory_config attribute in TTNN dialect:

tt-mlir/include/ttmlir/Dialect/TTNN/IR/TTNNOps.td

Lines 864 to 876 in 48ade5e

	def TTNN_ReshapeOp : TTNN_Op<"reshape"> {
	let summary = "Reshape op.";
	let description = [{
	Reshape tensor.
	}];
	let arguments = (ins AnyRankedTensor:$input,
	I32ArrayAttr:$shape);
	let results = (outs AnyRankedTensor:$result);
	let hasVerifier = 1;
	}

Can you give us a priority list of ops that we should handle first before we handle this systematically?

[odjuricicTT]

@sdjordjevicTT Concat and reshape are blockers for now. Agreed that this should be systematically handled.