Add logic to keep the last N logs (say 100) but delete older ones beyond that. The goal is to prevent disk space from getting used up by log files

This regression is associated with the websocket reconnect feature I recently added.

In the latest dev build (photonvision-dev-dev-13-g66a5fa8.jar), the Brightness slider in the Settings page doesn't dim a Gloworm's LEDs properly. If the slider is set to 100, the LEDs are fully on. If the slider is set to any other value, the LEDs are fully off.

https://discord.com/channels/176186766946992128/528555967827148801/740329176246255720

https://cdn.discordapp.com/attachments/725846784131203222/763161108852375582/2020-10-06_18-09-31.mp4

To reproduce:

• Set target grouping to dual
• Reload web interface

Current workaround is to manually run git tag dev

This can lead to connecting to a camera on a path it no longer is on.

This is for sure present in server mode, not sure about gloworm in client mode on robot

The install script given in scripts/install.sh fails when running during the "Downloading latest stable release of PhotonVision..." portion of the install script. This can be seen by curling only the link "https://api.github.com/repos/photonvision/photonvision/releases/latest" or by pasting it directly into the browser.

This failure is silent and the script says that the install was successful as there is no error checking.

Reproduce: set pipelineIndex in NetworkTables to -3.
Behavior with -1 or -2 will change to internal pipelines and should not do allow that either.

Save the latest calibration images inside the settings directory so the user can later reference them.

See attached. It seems to work fine at whatever resolution is already selected by default, but UI and modals don't show up with correct text:

Procedure:

Start calibration, take a 25 snapshots, then refresh the page.

Snapshot count goes back to 0 and UI button returns to "cancel calibration".

Clicking "Cancel" actually starts to perform the calibration

Taking an additional snapshot does reset the GUI state. I think snapshot count just needs to be sent as part of the vue data sent on initial page load when a calibration is in process?

https://cdn.discordapp.com/attachments/725846784131203222/794345048857772032/unknown.png

Because image resizing is currently done by the frame consumer, and its not updated when pipelines are switched, stream resolution doesn't change with pipes. https://cdn.discordapp.com/attachments/725846807019782174/784190774400122880/2020-12-03_17-53-10.mp4

This should not be the default

Since images can be saved at runtime, and log files can get larger, we should alert users when their disk usage is growing too large.

A new camera that is at the same path as a previously connected camera will match to the previous camera config, even if the base name is different.

The camera shutter speed seems to be 0 (or very small) when the exposure slider is set to 0, and some fixed value when the slider is greater than 1.

The shutter should be able to be set from 1 to 6,000,000 microseconds (or some subset of that range).

See this GIF:

https://cdn.discordapp.com/attachments/725846784131203222/759174444659900456/PhotonVision_-_Mozilla_Firefox_2020-09-25_18-07-18_Trim.mp4

Right now, there's a requirement that PnPTab.vue and TargetModel.java are manually kept in sync to ensure the user has the ability to select from any of the available target models.

Update the backend transmit a list of nice human-readable strings naming the target models to the frontend, and use them to populate the correct dropdowns.