HVG always / never + CI update

[pcm32]

Description

This PR allows to bias highly variable gene results for use cases where to want to ensure that certain genes are considered or are not considered highly variable genes (ie. there are cases where you don't want Igg* genes for instance to control the clustering and other processes downstream that depend on HVGs).

I took the chance of updating the CI to the latest Tools IUC, with some minor modifications.

Uses main instead of develop from now on.

Type of change

• [ ] Bug fix (non-breaking change which fixes an issue)
• [x] New feature (non-breaking change which adds functionality)
• [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
• [ ] This change requires a documentation update

Checklist

• [ ] I have made any required changes to upstream dependencies for a tool wrapper, and they are available in distribution channels (e.g. Pip, Conda).
• [ ] If I have updated the underlying software for a tool wrapper (e.g. scanpy-scripts by changing the value of @TOOL_VERSION@), then I have reset all 'build' values to 0 (e.g. @TOOL_VERSION@+galaxy0)
• [x] If I have updated a tool wrapper without a software change, then I have bumped the associated 'build' values (e.g. @TOOL_VERSION@+galaxy0 @TOOL_VERSION@+galaxy1). It is acceptable to do this as well when the cli version changed but not the underlying tool (to avoid issues in the coming point).
• [ ] If I changed the version, the @TOOL_VERSION@ part of the version does not contain any + symbols within, otherwise this will break tool ordering on the interface and the default tool being picked. Tool version should always conform to PEP440 to avoid this issue. The only + should be the one preceding galaxy<build> (unless that all the versions from that tool previously followed a different pattern).

[pcm32]

Merging this as previously approved by @anilthanki on #319 .