Caper

Fixed `PyYAML` installation issue

- Use file metadata DB by default (`--db file`). If you see any DB connection timeout then switch back to `in-memory` DB with `--db in-memory`, then you will pass the error but call-caching will be disabled.
- Changed default file path for such file DB (`--file-db`): `${local_out_dir}/caper-db_[WDL_BASENAME]_[INPUT_JSON_BASENAME]`. `local_out_dir` can be defined either in command line or conf file (`~/.caper/default.conf`).

- Can use a relative path in input JSON. If it still doesn't work. `pip install autouri --upgrade`

- Changed local file hashing strategy from `path+modtime` to `xxh64` to prevent call-caching's cache miss due to the same text file with different `modtime` (e.g. modified by `touch`).

- Fixed `filelock` issue. Run `pip install autouri --upgrade` to make sure to have `autouri>=0.4.4`.

- `caper init` makes a simpler configuration file `~/.caper/default.conf` with less commented instruction and parameters;
- Made documentation simpler.

- Upgraded Cromwell from 65 to 82.
- Fixed issues
- Issue 177: SGE/PBS/LSF resource parameters were not correctly propagated to `backend.conf`.
- Issue 180: GCS `engine` stanza is missing in `backend.conf`

Fixed issue 177
- Some parameters in CLI (`sge_resource_param`, `pbs_resource_param` and `lsf_resource_param`) were not passed correctly to `CaperRunner` class constructor.

Deprecated `scg` and `sherlock` backends.
Use `slurm` backend instead and follow instruction on Caper's conf `~/.caper/default.conf` after `caper init slurm`.

New HPC feature
- Unified command `caper hpc` for all HPC backend types (`slurm`, `sge`, `pbs` and `lsf`).
- `caper hpc submit WDL -i INPUT.json --leader-job-name GOOD_NAME_FOR_LEADER [--conda | --singularity]`
- `caper hpc list`: shows all Caper leader jobs
- `caper hpc abort JOB_ID`: terminate a Caper leader job including child jobs

Default walltime for a leader job is 48 hours by default. Check Caper's conf `~/.caper/default.conf` to change it.