Elastic-blast

Bug fixes:
* Fix bug that leads to invalid GKE cluster names

Bug fixes:

* Fix bug that causes ElasticBLAST on AWS to fail to read from S3

Improvements:
* Improve support for multiple local SSDs on AWS
* Consistent tagging of AWS EC2 Launch Template
* Machine selection in AWS is now stable

Bug fixes:
* User names with certain characters do not result in cloud provider errors any more

New features:
* Provide support for BLASTDB metadata files version 1.2
* Add support for `--exit-code` option in `elastic-blast status`
* Check for resource files in PEX
* Provide scripts to check for undeleted ElasticBLAST searches

Improvements:
* `elastic-blast status --verbose` prints BLAST+ error message, if a search failed due to BLAST errors
* Additional checks for num-cpus parameter that prevent unschedulable search jobs
* Improved error message for BLAST database initialization with local SSD in GCP
* Cluster name can be set via cluster.name configuration parameter
* Incorporate mt_mode choice when selecting machine-type
* `created` tag is recorded in UTC time

Bug fixes:
* User names with certain characters do not result in cloud provider errors any more
* Limit the number of concurrent `aws s3` invocations when downloading BLASTDB to prevent OOM
* Initialize local SSD before tagging to ensure they are ready to use in spite of tagging errors

New features

- Provide scripts to create ElasticBLAST janitor role in AWS
- Search for canonical ElasticBLAST janitor role in AWS to enable its functionality
- Validate CSP region

Improvements

- Improve error message when ElasticBLAST Janitor in GCP cannot run due to missing permissions
- Provide error messages when invalid BLAST+ options are specified
- Disable auto-upgrade in GKE clusters
- Improved batch-len defaults
- Restore janitor in AWS

Bug fixes

- Additional fix for GCP disk leak: wait for PV to unmount when jobs are still running

New features

- Job submission on the cloud for GCP
- Auto-shutdown for AWS
- Add programmatic SSD quota check

Improvements

- elastic-blast automatically sets memory-limit(s) by default
- Make cleanup-stale-gcp-resources.py public
- Change batch-length on the fly if number of query batches is less than number possible BLAST parallel jobs
- Run summary retrieves paginated results, tracks all runtimes, speed collection improved (letters/sec/CPU)

Bug fixes

- Autoscaling while job submission is taking place may lead to kubectl disconnects