`runner.sh` Parses Some Files Incorrectly

Describe the bug
The runner.sh script parses out the output from jupytext into something for feeding into pep8. sed is used in places to pick parts out, however for some weirdly formatting Python this will introduce syntax errors. For example, the following will not be interpreted correctly:

if epoch % 10 == 0:
        print(
            "Train: reconstruction: %.4f; generator: %.4f; discriminator: %.4f; perceptual: %.4f; kld: %.4f"
            % (
                epoch_recon_losses[-1],
                epoch_gen_losses[-1],
                epoch_disc_losses[-1],
                epoch_perc_losses[-1],
                epoch_kld_losses[-1],
            )
        )

The solution is to reformulate the code to avoid these sorts of issues. This example here isn't very attractive code anyway so a refactor to break the contents into multiple lines which are read properly wouldn't hurt anyhow. There will be other cases potentially that are well-formed code but still introduce parsing issues.

This is a minor issue I haven't had time to work on a solution for other than to refactor code, a solution to using jupytext and sed better should be investigated later.

To Reproduce
This can be seen as a problem in PR #1794 currently though a fix is being worked on.

Expected behavior
Script shouldn't introduce issues

Start by inspecting runner.sh and its jupytext and sed extraction path, then compare the behavior described in PR #1794. Done means the script preserves well-formed Python while extracting content for pep8 without introducing syntax errors, including the multiline example shown here.