Add gitattributes file to adjust linguist settings (#13977)

Github uses a ruby library linguist [1] for detecting the languages used
in a github repository. The detection this library does is not perfect
as it relies on file extensions and obviously it doesn't know about
context around certain files. Linguist provides tools for correcting
these issues so the reported split of languages in a repository is more
accurate. This commit adds a .gitattributes file to Qiskit to adjust how
linguist interprets some files to provide a better measurement of the
languages used in the repository. This involves three main changes:

1. Marking the qasm files in test/benchmarks/qasm as generated code as
   they're all generated output for measuring qasm parsing time or to
   have a deterministic circuit definition. These aren't source files
   that should be counted as part of the language breakdown.
2. Marking the visualization text references in LaTeX and txt files as
   generated, as they're generated from the visualization output to
   serve as a comparison for testing.
3. Marking the qasm libraries using the *inc extension as OpenQASM not
   C++.

With these three changes the languages used by qiskit is a better
reflection of the current state of the repository.

[1] https://github.com/github-linguist/linguist

Author: mtreinish

.gitattributes

@@ -0,0 +1,6 @@
+# Exclude benchmark qasm and reference text:
+test/benchmarks/qasm/** linguist-generated
+test/python/visualization/references/*.tex linguist-generated
+test/python/visualization/references/*.txt linguist-generated
+# Mark qasm libs as qasm not c++
+*.inc linguist-language=OpenQASM