Tainted instruction plugin optimizations.

[MarkMankins]

• Made globals static.
• Added "suppress_redundant_taint_reports" option. -- When enabled, won't output the same PC multiple times consecutively. -- Only used when pandalog is not enabled.
• In the for loop that computes num_tainted in taint_change: -- Exit early if not pandalogging.
  -- Further looping in this case isn't necessary as output doesn't change.
• Removed redundant ifs.
• Fix memory leak (ti->taint_query allocated via malloc was not freed.)
• Save "pc = 0x..." generated string in case it can be reused.
• Moved assignments to last_asid and last_pc to avoid unnecessary assignments.

[AndrewFasano]

CI is a bit broken right now, but this looks good, thanks!

[AndrewFasano]

@coolkingcole I don't think we've seen this CI issue before - It was trying to push a new container on the PR and getting permission denied. The fact that it didn't have permissions seems good, but it probably shouldn't have tried or reported a failure about it.

[coolkingcole]

I believe this is a user permissions error related to to role in the org. We should add additional branch protection to avoid malicious workflows. For now we have the setting Fork pull request workflows from outside collaborators set to the most restrictive setting which restricts Outside Collaborators, which @MarkMankins is.

[AndrewFasano]

@coolkingcole makes sense that we'd restrict roles - but if someone with more permissions opened a PR, we wouldn't want the PR itself to publish a new release. We'd only want that after the PR is merged, right?

[coolkingcole]

If we want that, we could change it to something like this

on:
  push:
    branches:
      - dev
      - stable
  pull_request:
    types:
      - closed