Clean, Streamline, Debug

[Lauenburg]

PR #60, #63, #65, and #69 introduced fundamental changes to SynAnno's core. To ensure a smooth operation, the code requires the straightening out of some computational inefficiencies, optimizing the frontend layout, and intense testing/debugging.
This issue is a collection of such tasks.

Problem Statement

1. Processing button - Neuron-centric approach
   If the user does not restrict the post- and pre-synaptic IDs, the tool can end up with 10k pages. Thus the user would never reach the end of the pages, and thus, the "Error Processing" button that is so far only available on the last page.

2. Benchmark and optimize the page vise loading for the neuron view
   The loading of the 18 instances per page currently takes about a minute. This means that if a user wants to prove read 180 instances that get split over 10 pages, he will have to wait 10 min in total for loading. The loading time for the view-centered approach is significantly less.

3. Clean up and redesign the "Loading Data" page
   We fully transitioned to the view and neuron-centric loading shema. The loading view still contains stale components, and the logic when which buttons get enabled is not sound.
   -Missing loading screen when saving/processing FN

4. Inverse Depiction of EM in tile view.
   The colors of the EM crops seem to be inverted.

5. NG does not display data
   Currently, the NG instance does not display the EM or seg. data.

6. Clean up stale code
   The tool was fundamentally changed, and we are likely left stale, legacy code that could be cleaned up

7. Error Handling
   Currently, under certain configurations, the errors are not smoothly handled:

• Not providing the relevant bucket secret

Expected Functionality

1. Processing button - Neuron-centric approach
   The user should be able to at any time assign the appropriate error labels and download the information for the instance that he has so far annotated.

2. Benchmark and optimize the page vise loading for the neuron view
   The loading time should be more similar to that of the view-centered approach and significantly less than it currently is.

3. Clean up and redesign the "Loading Data" page
   Clean layout. Buttons that are enabled when every necessary requirement is given.

4. Inverse Depiction of EM in tile view.
   The colors should show correctly.

5. NG does not display data
   When clicking the 'View in NG' button, the user should be able to view the EM and seg. data.

6. Clean up stale code
   Clean code.

7. Error Handling
   Catching the error and returning a helpful description.

Proposed Solution

1. Processing button - Neuron-centric approach

• Provide a button that lets the user finish his work no matter which page he is on.
• For the neuron-centric approach, the user should be able to provide a JSON that holds the information for the n first instances where n < N and N is the complete number of synapses associated with the provided IDs.

2. Benchmark and optimize the page vise loading for the neuron view
   Benchmark the individual steps, find the inefficiency, and optimize them if possible.
   Try out the parallel loading functionality from CloudVolume.

3. Clean up and redesign the "Loading Data" page
   Clean layout. Buttons that are enabled when every necessary requirement is given.

4. Inverse Depiction of EM in tile view.
   View the EM data in the demo Neuroglancer. Check where the colors get inverted.

5. NG does not display data
   Assess the NG version and identify reasons behind incorrect data display.

6. Clean up stale code
   Clean code.

7. Error Handling
   Add try and excepts where relevant.

[Lauenburg]

Resolved the Neuroglancer depiction issue for the view and neuron mode with commit 34868423df5fae8bdcf2c633fc4a7223a8ca8b68 and efcb50e45caa0d744497d7134a5223278619dbe4 respectively.

[Lauenburg]

Resolved the NG color inversion with commit 97755283f455ebd8c6b420bd74ddeb33313e414c

[Lauenburg]

Optimized loading times in neuron mode with commit efcb50e45caa0d744497d7134a5223278619dbe4 by implementing multithreading to handle instance sub-volume downloads, an I/O bound process. This enhancement significantly improved performance, reducing loading times by a factor of 6-7.