Off-by-one error in longest strike features

[blu3r4y]

The two features longest_strike_below_mean and longest_strike_above_mean are implemented in a "less-than" and "greater-than" fashion, which yields incorrect results.

Here is a MWE where µ = 3 and the longest strike below and above is 2 - tsfresh yields 3.

assert fc.longest_strike_below_mean([1, 2, 3, 4, 5]) == 2
assert fc.longest_strike_above_mean([1, 2, 3, 4, 5]) == 2