Add min/max value constraints to score and confidence slots.

[gouttegd]

Resolves #372

• docs/ have been added/updated if necessary (not applicable; the doc was already ahead of the model, this PR is about updating the model to align it with the doc)
• make test has been run locally
• tests have been added/updated (not applicable)
• CHANGELOG.md has been updated.

The description of the confidence, semantic_similarity_score, and similarity_score slots says that those slots expect a value between 0 and 1, but the model does not enforce that. LinkML provides a way to formalise this kind of value constraints (which could then be automatically enforced by a runtime that is aware of those constraints), so we do exactly that here: we add minimum_value and maximum_value constraints to the aforementioned slots.

[matentzn]

We probably need a better SOP for testing these kinds of PR. I had been working on a pipeline, but the truth is, what would actually need to happen is:

1. Build LinkML model (this ensures that the model is syntactically well-defined)
2. Build version of sssom py with that locally cached linkml model
3. Run sssom-py test suites

Else I dont know if a change like this will break existing tests. Did you test these changes somehow?

[gouttegd]

1. Build LinkML model (this ensures that the model is syntactically well-defined)

I believe the existing test suite here (the one that is run by make test) takes care of validating the model at least from a syntactical standpoint. First, because what would even be the point of the test suite if it did not even do that, and second, because when preparing this PR I initially and erroneously wrote maximal_value instead of maximum_value, and the test suite (which I ran locally before even committing any changes) correctly rebuked me for that.

2. Build version of sssom py with that locally cached linkml model
3. Run sssom-py test suites

Why would we want sssom to fail because of problems that could exist in sssom-py?

If SSSOM-Py breaks when the model enforces a constraint that has always been a part of the spec but was not previously enforced, then it's a problem with SSSOM-Py, not with the spec.

[matentzn]

I believe the existing test suite here (the one that is run by make test) takes care of validating the model at least from a syntactical standpoint.

You are right, forgot about that; it already does what I thought it did not.