[Example] API for iterating layer outputs

[pinzhenx]

Fix #511

[pinzhenx]

Select model in UI, then issue the following command in console

// iterate all layers with one backend
utils.iterateLayers([{backend: 'WebML', prefer: 'sustained'}]);

// iterate specified layers
utils.iterateLayers([{backend: 'WASM'}], [1, 3, 5]);

// iterate all layers with multiple backends
// take the first backend's result as reference and compute variance of the rest with it
utils.iterateLayers([{backend: 'WASM'}, {backend: 'WebGL'}]);

[BruceDai]

We have verified this PR.
@huningxin Please help to merge it, thanks

[pinzhenx]

@huningxin Conflicts resolved. Please have a look, Thanks.

[pinzhenx]

@zhaoming0 @fujunwei Layer iterating tool is now available publicly. The usage is shown above. It’s an example level debugging tool and now available in Image Classification and Deeplab. I hope you find this useful.

[fujunwei]

@pinzhenx thanks for providing the useful tool.