Improve plotting code - robustness, unification, remove code repetition

[keiran-rowell-unsw]

Plotting refactor, drops duplicate files, ~300 lines of repetition beyond files, and reads more cleanly.
Implements nf-core#484, closes nf-core#485, and inches towards nf-core#373.
Will verify this implements nf-core#363

• HTML report now drops sections where the metric isn't available ( e.g. ESMFold)
• One generate_report.py, one report_template.html, all plotting logic in plot_utils.py; conceptually separates out the process constructing a report from plotting functions that use plain .tsvs
• Uses a sort_structures_by_rank() utility in plot_utils.py to provide a place for program-specific ranking rather than use a matcher =~ with regex in the .nf pipeline.
• Each program sorting logic can be expanded upon in the util.
• Injection of content in report_template.html now cleaner
• Everything as plotly objects, no need to write-our and read-in .png files, users can save from the HTML report from plotly toolbar if they want to
• post_processing.nf massively simplified due to removing program-specific exceptions
• modeChannel() used consistently, cleaning many lines of adding in mode
• ch_multiqc now cleared, was previously used to pass along plddt which is now simply extracted from structures with biopython, and makes way for a bulk folding report in Bulk MultiQC reporting as a pipeline-installed Python module nf-core/proteinfold#450.
• Switches to unpkg.com as a more robust and up-to-date npm CDN. rawgit sunsetted in 2019

TODO:

• Make a plotly sequence coverage object look as nice as the prior .png

TESTING:

• In draft mode currently because I haven't run on all modes or validated comparison reports fully