Add history-based autocomplete

[remulasce]

Autocomplete dropdown should show recent/high-use stops immediately on focus when no text has been entered. These suggestions should receive higher priority when 'regular' suggestions appear They should also be filtered by the text entered, as with normal suggestions.

[remulasce]

In progress. Tuning on my device.

[remulasce]

It's mostly done, including text-filter.

Needs a little tuning on the total-number-used though. Currently once you use any selection 5 times, it maxes out that priority calculation. When really you might expect to use the same stop every day for months, and you could add the same stop 5 times if you open/close the app 5 times to see when to leave for the first train, then wait for the station, and set it up all back to get home the same day.

[remulasce]

Final tuning:

• I want daily-use stops to come out much ahead of 1-10 time use
• I want stop priority to decay to 0 levels over time

Solutions:

• Total use number gets you dailies > else
• Occasional decay on number gets you towards 0
• ???
• Profit
• Recency still buys you a weekend of looking-it-up

So let's do total use with occasional decay.

[remulasce]

Decay options:

• Time
• Internal app use

Time seems nice if you're using it every day. It's a constant factor that I can tune to my expectations. But It takes more to implement.

By use (say, anytime anything gets selected) still meets the criteria. But if the user isn't using the app a lot, his predictions will remain from old, not-actually-relevant trips.

Let's go with internal app usages- so every time user selects any stop, other stops get a little kick. I don't want to optimize for such long-away cases for users who barely use the app.

[remulasce]

Permadecay implemented.

Ready for wider testing (like actually using the darn thing.)

[remulasce]

Couple days' use, looks great.