Add missing build methods to the RGBD and SpatialDetectionNetwork nodes

[lnotspotl]

Purpose

Add missing build methods to the RGBD and SpatialDetectionNetwork nodes.

CU task: https://app.clickup.com/t/86c6kyk13

[moratom]

I was thinking we could add a std::variant<dai::node::StereoDepth, dai::node::NeuralDepth, dai::node::Tof> depthinput as the input to the functions, to avoid too much duplication.

Would this be an option @lnotspotl ?

• we should add a test for the NeuralDepth overloads, maybe holistic replay can be used, similar to what we have for NeuralDepth tests.

[lnotspotl]

@moratom I've added a typedef for a variant that encapsulates all three depth map sources: StereoDepth, NeuralDepth and finally ToF

I've also updated two of the examples such that they accept a depthSource flag that can be set to "stereo", "neural" or "tof" such that one can experiment with all three of them.

I still need to perform testing, make sure all possible configurations work.

[lnotspotl]

Had to remove support for "tof" from the SpatialDetectionNetwork example as the camera was running out of memory.

[lnotspotl]

The difference was in the synchronization node's settings. In case of the autocreate feature, synchronization would happen on host. This was not the case for the non-autocreate example where synchronization was happening on device.

The two examples now essentially work the same.