Add compatibility fixes for pandas 3 and torch 2.10

[stes]

• fix readonly-issue in pandas, from 3.0.0 onwards Make CEBRA compatible with pandas 3.0.0 #287
• upgrade build pipeline to latest torch -> errors/deprecation warnings still need to be resolved, see Upgrade to torch 2.10, resolve deprecation warnings #288, not part of this commit
• if readonly arrays are passed, an informative error message is passed

Fix #287

[stes]

Added stricter pytest checks. If torch raises a warning because we pass a non-writable tensor to CEBRA, this will be treatet as an error for the tests we are running. Where applicable, we instead add a custom warning from CEBRA.

[MMathisLab]

Lgtm 🚀