Relax python restrictions to allow Python >= 3.8 -> 3.9

[oschwengers]

Hi Savojardo, currently, DeepSig pinpoints dependencies for Python, keras and tensorflow to 3.8, 2.4.3 and 2.2.0, respectively. However, this will forestall many useful updates and might hinder the creation of larger environments using DeepSig with other tools.

To integrate DeepSig into Bakta, I once again took a look and successfully managed to increase Python, keras and tensorflow dependencies to 3.9, 2.6.0 and 2.6.2, respectively. This also worked with Python 3.8:

• https://github.com/BolognaBiocomp/deepsig/commit/a7b0e64bb2879b545eed2b8f16271761b260e1ed
• https://github.com/BolognaBiocomp/deepsig/commit/f6a182394048abc7833188143e4750eb1c854ab0

To streamline these verison bumps I also added two unit tests that are automatically triggered on each commit and executed by the GitHub CI system:

• Python packaging: https://github.com/BolognaBiocomp/deepsig/commit/7198f0ba0cb719a7b67332b42e1216e0a1a63ad1
• Conda based execution of DeepSig on a very tiny test data set (2 bacterial aa sequences): https://github.com/BolognaBiocomp/deepsig/commit/0ba2842225ba7dc1ffe85c124323046d2aca5793

I think this might be a worthy step forwards in terms of long term software stability and usability. Could you please take a look at this, test it yourself and consider this PR?

[oschwengers]

You can take a look at the the last and successful unit tests at:

• https://github.com/oschwengers/deepsig/actions/runs/1717955509
• https://github.com/oschwengers/deepsig/actions/runs/1717955510

Unfortunately, I had to disable these tests from MacOS as the CI was not able to execute biocrf-static. I guess this is not related to the CI and maybe a distinct issue with DeepSig and MacOS.