docs: full task compliance with style-guide

[a-frantz]

I believe all task meta/parameter meta should be fully spec compliant now. Except the ngsderive file which is being handled in #108 . If any tasks are noticed with non-compliant meta sections, please bring to my attention. (Sure would be nice to have a tool that would verify this for us)

And a new style rule! All "deprecated" tasks should be at the bottom of the file. I figure that's an uncontroversial rule.

While working through this, I also made some updates to tasks, mainly exposing more parameters in samtools.wdl.

Also set a new soft limit for "least amount of memory to allocate a task", which is 4gb. Implementing this meant lowering a few tasks' default allocation from 5gb to 4gb for consistency. No allocations were raised (as none were less than 4).

Still TODO in other PRs:

• meta sections are currently missing from all our workflows
  • hopefully the outputs section will be easier to write now, as we can just copy+paste from the task outputs sections
• Headers are still inconsistent, and to boot I don't think a single one conforms to what we "enforce" in the style-guide.
• I added a whole bunch of small TODO comments that I didn't feel like addressing at the time. They can be dealt with in another PR
• Investigate whether we can get away with a lower memory allocation of 3gb. This would allow running on most of the smallest cloud instances.

[a-frantz]

New repository rule! Beginning with this PR, ALL tasks must be covered by at least one test and all tests must be passing (unless there's good reason for a test to be failing). No CI has been implemented yet, so we're relying on ad-hoc testing and self-reporting of success.

I have added tests that were missing, and have gotten everything passing. EXCEPT:

• the cellranger tasks were not validated. It's a pain to configure my cluster singularity environment to test them, and we've discussed moving the cellranger files to their own repository. AKA I'm treating their tests as OOS.
• the qc_summary test is also failing. The underlying task is out of sync with the multiqc task, and qc_summary makes bad assumptions about what fields will be present. It needs to be re-written from scratch. That will likely be the final task before the QCv2 official release. It's a wasted effort fixing it till that later stage.