Add new Lint/FloatComparison cop

[fatkodima]

This is a known fact that comparing floats is bad, so this cop checks for the presence of precise comparison of floating point numbers.

a = (0.3 * 3) + 0.1 # 0.9999999999999999
b = 1.0

a == b # false 

# bad
x == 0.1
x != 0.1

# good
(x - 0.1).abs < Float::EPSILON

# good
tolerance = 0.0001
(x - 0.1).abs < tolerance

Reference: https://www.jetbrains.com/help/resharper/CompareOfFloatsByEqualityOperator.html

[bbatsov]

Great cop idea! Can you also submit a PR for this to the Ruby style guide?

[fatkodima]

Updated with all suggestions.

[bbatsov]

Thanks!

[dvandersluis]

Should this flag comparisons to 0 or 0.0? I don't believe there are precision issues with zero (Float#zero? is true for 0.0).

See: https://github.com/ruby/ruby/blob/c87043283ef7114db54857ae13798e3d7cb031c0/numeric.c#L1743-L1747 and https://github.com/ruby/ruby/blob/1035a3b202ee86bf2b0a1d00eefcfff0d7ab9f6b/internal/numeric.h#L137-L141