Quality estimation based on GPT and Llama3, integrated with InstantTranslate and translate5 editor

Please ask Marc upfront, to enhance the issue description.

So far it looks like:

If active: All segments of a task or the InstantTranslate text field should be send to a quality estimation to the LLM in a way, that the LLM can respect the context of the segments.

The LLM should estimate the quality of the all over text AND the quality of each segment.

The overall quality should be displayed in the analysis of the task and in InstantTranslate below the text field and next to the file available for download.

It should be configureable in translate5s config, at what quality level the values should be highlighted green, yellow and red.

The segment based quality should be displayed in InstantTranslate text field in a mouse over tooltip of the segments including the colored highlight (if there are mulitple segments in the result). If there is only one segment, it should be displayed below the segment, again with the background color as defined in the config.

In InstantTranslate text field it is important, that the quality estimation does not slow down the showing of the result. It should be requested in a separate request and as long as it did not respond, the place that will show the quality estimation result will show a turning wheel.

In translate5's editor there will be a new column "Quality estimation" next to the match rate column. It is only available, if quality estimation is active for the task. If yes, it shows the estimation value colored by the above mentioned background color. If the user edits and saves the segment, the quality estimation of this segment is redone and afterwards the value for the segment is adjusted. It is send to the LLM including the context of the 3 previous and following segments.