Detection algorithm rework

This version contains a complete rework of the detection algorithm. I have been more rigorous with my test protocol lately so I have been able to measure the improvements.

The protocol is the following:

• 3 images (from my OnePlus 10 Pro device, 3216 x 1440): one small (the size of a launcher icon), one medium (the size of 4 launcher icons in width, 1 in height), one big (the width and height of a 4x4 grid of launcher icons
• Each of this image is used in 2 conditions, on detecting at position, one on the whole screen
• Detection quality to maximum (1200)
• Detect during 10000 screen frames

Here are the results:

• RAM usage went from 332Mb max to 161Mb max
• Success went from 8451 to 10000 (100% of correct detections)
• Average detection time is divided by 10 for EXACT detection, and by 5 for WHOLE SCREEN.

As you can see, the results are really great, and the improvements should be seen by the user. As a consequence, i have updated the Detection Quality feature, by increasing its minimum to 800, and the maximum to 3216 (which is, according to my tests, still faster than the old maximum). All your scenario detection qualities will be automatically updated and increased in order to improve automatically all your scenario.

Condition operator and click on condition

This feature wasn't really clear, you never know on which condition it will click. This is now fixed by adding a condition selector to the click action in the case the event condition operator is set to All conditions.

Tutorial update

A new tutorial has been added, and all of them have been translated in the supported languages.

Stabilization

A lot of corner cases have been fixed in this version, and the detection algorithm revamp should also be more robust.
The user crash free rate on the 2.1.4 version is 99.76%, let's try to reach 99.9% !

Note

An unexpected issue has happen to the update apk. The apk linked with this release post is the 2.3.1