Improve season uncertainty reset by considering players who have not played for a while

[jauggy]

For the past uncertainty reset, everyone's uncertainty was set to
max(5, current_uncertainty)

This PR will make the uncertainty reset more dynamic, by giving larger reset for players who have not played in a while.

It will reset ratings to max(target_uncertainty, current_uncertainty)
where target_uncertainty is:

• 5 for players who have played within one month
• 8.333 or default uncertainty for players who have not played for a year or more
• Something in between these two values for players who have not played for more than one month and less than one year.

To calculate the in between value, we use linear interpolation using the formula: https://www.cuemath.com/linear-interpolation-formula/

We can see the values would be something like this:

    one_month = 365 / 12

    result = SeasonalUncertaintyResetTask.calculate_target_uncertainty(one_month)
    assert result == 5

    result = SeasonalUncertaintyResetTask.calculate_target_uncertainty(one_month * 2)
    assert result == 5.303030303030303

    result = SeasonalUncertaintyResetTask.calculate_target_uncertainty(one_month * 3)
    assert result == 5.6060606060606063

    result = SeasonalUncertaintyResetTask.calculate_target_uncertainty(one_month * 6)
    assert result == 6.515151515151516

    result = SeasonalUncertaintyResetTask.calculate_target_uncertainty(one_month * 9)
    assert result == 7.424242424242426

    result = SeasonalUncertaintyResetTask.calculate_target_uncertainty(one_month * 11)
    assert result == 8.030303030303031

    result = SeasonalUncertaintyResetTask.calculate_target_uncertainty(365)
    assert result == 8.333333333333334

Testing

1. Open up the database
2. View teiserver_account_ratings table and check the last_updated column
3. Run iex -S mix phx.server
4. Run Teiserver.Battle.SeasonalUncertaintyResetTask.perform()
5. This should update the uncertainties and also the last_updated column from step 2.

[L-e-x-o-n]

Nice improvements.
What do you think about min_uncertainity as a parameter passed to the task, in case there is a time in the future where a smaller (or larger) uncertainty reset might be wanted?

[jauggy]

OK Lex I will make a 5 a variable that gets passed in. Marking PR as draft until I add it in.

[jauggy]

Closing in favour of #540
Trying to do this in sql