[Discussion]: Lower version bounds of anytree and imageio? - conda and pip conflict

[Saransh-cpp]

2 of our dependencies have different latest versions on conda and pip, which created a minor issue in the conda release. Lowering the version bounds for these dependencies to a common value worked.

---

The workflows failed when I update the versions to match our PyPI requirements in conda-forge/pybamm-feedstock@cedf7b6, so it looks like the workflows are not installing the dependencies from pip -

The following packages are incompatible
├─ anytree >=2.12.0  does not exist (perhaps a typo or a missing channel);
└─ imageio >=2.32.0  does not exist (perhaps a typo or a missing channel).

Originally posted by @Saransh-cpp in conda-forge/pybamm-feedstock#54 (comment)

I think we should lower the bounds for these dependencies in the next PyPI release as well - to maintain uniformity between the conda and PyPI packages - but I'm not very sure of this.

Originally posted by @Saransh-cpp in conda-forge/pybamm-feedstock#54 (comment)

[agriyakhetarpal]

I read up a bit on this and imo lowering the version bounds should be a good idea, especially for packages that are not as big or as frequently maintained compared to those in the Scientific Python ecosystem. Packages that are outside of the ecosystem and subsequently those with lesser adoption or usage rates generally do not feel the need to bring conda-forge releases at the same time as their PyPI release counterparts, which brings us to this version mismatch. Out of our core dependencies: numpy and scipy work in tandem, xarray is quick at releases, but casadi and anytree tend to be on the slower side; their versions should be kept in sync from PyBaMM 24.1 onwards.