Ensure csr matrices are in "canonical format" in Hazard objects

[peanutfun]

Changes proposed in this PR:

• Use getter/setter methods and private attributes to ensure that any csr matrix passed into a hazard is in "canonical format". This ensures that the csr_matrix.data attribute reflects the exact values present in the matrix.
• Update tests

This PR fixes inconsistencies in the csr matrix reported in #891, but does not fix the problem of unintentionally summed values (it just ensures this problem is visible when looking at the matrix data).

PR Author Checklist

• [x] Read the Contribution Guide
• [x] Correct target branch selected (if unsure, select develop)
• [x] Descriptive pull request title added
• [x] Source branch up-to-date with target branch
• [x] Documentation updated
• [x] Tests updated
• [ ] Tests passing
• [x] No new linter issues
• [x] Changelog updated

PR Reviewer Checklist

• [ ] Read the Contribution Guide
• [ ] CLIMADA Reviewer Checklist passed
• [ ] Tests passing
• [ ] No new linter issues

[peanutfun]

Petals compatibility tests are failing because we sometimes assign LIL matrices there instead of CSR matrices. This can be fixed before or after merging this PR.

[peanutfun]

Petals compatibility is addressed here: CLIMADA-project/climada_petals#129

[peanutfun]

@chahank What do you think about this updated version? I still have to adapt the test.

[peanutfun]

@chahank Waiting for your responses to move forward 😇 👉 👈