Expand bug data as predictors of vulnerabilities

[andymeneely]

Let's go over the labels again and identify types of bugs we are particularly interested in. We can repeat the ideas of pre_ and post_ analysis of #189, and maybe get some results like "Fixes for performance bugs early can lead to security bugs later".

To close this issue, let's decide on a list of labels and metrics.

[andymeneely]

Here's an interesting query:

Label.joins(:bug_labels).group('labels.label').select('labels.label,count(*) as cnt').order('cnt DESC')

[andymeneely]

Labels we care about:

• type-feature
• type-compat
• type-bug-regression
• type-bug-security

Labels we want to look into as possibilities:

• cr-tests-fails
• stability-crash
• bulkmove
• build
• needs-reduction
• has-reduction

[andymeneely]

Based on that, let's go with these pre_ bugs:

• num_pre_features
• num_pre_compatibility_bugs
• num_pre_regression_bugs - which is now based on type-bug-regression label
• num_pre_security_bugs
• num_pre_tests_fails_bugs
• num_pre_stability_crash_bugs
• num_pre_build_bugs