Hidden directories containing a large number of files impacts performance

[dark-panda]

Finding hidden files can have a performance impact when there are hidden
directories in the working directory that contain many files, as the
hidden_files array can become rather large, resulting in a long
start-up time while the target files are determined. Using a binary search
can cut down on this start-up time considerably.

As an example, on my machine, I have a .bundle directory containing my local
bundle, which contains tens of thousands of files. This causes the target_files
array to be quite large as a result. Scanning on these using Array#include?
becomes at worst O(n), while using a binary search will be on average O(log n).
This change reduced the start-up time on my Rubocop scans from around
100 seconds to around 15 seconds when analyzed using Benchmark.

Expected behavior

Quick start-up time.

Actual behavior

Potentially slow start-up time when a large number of hidden files are present.

Steps to reproduce the problem

Create a hidden directory containing a large number of files and run Rubocop.
The issue will worse as the number of files increases appreciably.

RuboCop version

1.6.1 (using Parser 2.7.2.0, rubocop-ast 1.3.0, running on ruby 2.7.2 x86_64-darwin18)
  - rubocop-performance 1.9.1
  - rubocop-rspec 2.0.1