refactor: add MissingDataError

[tristan-f-r]

While this only affects actual runtime code with an updated, consistent error message on malformed data, the goal of MissingDataError is to create a finer distinction between actual unexpected SPRAS errors and expected one, to further #21 (comment). Closes #393.

[read-the-docs-community]

Documentation build overview

📚 spras | 🛠️ Build #31038933 | 📁 Comparing e9fec74 against latest (de03d52)

🔍 Preview build

Show files changed (4 files in total): 📝 4 modified | ➕ 0 added | ➖ 0 deleted

File	Status
genindex.html	📝 modified
htcondor.html	📝 modified
fordevs/modules.html	📝 modified
fordevs/spras.html	📝 modified

[agitter]

Do we need to break the get_node_columns signature? Adding the checks within get_node_* functions makes sense to me, as does the new error type in general. My understanding is that the scope argument exists only so we can pass it through to the error message. That is more readable, but adds overhead in using these functions. The alternative would be to use the stack trace to see which algorithm generated the error.

Those are my initial thoughts before a deep review.

[tristan-f-r]

Agreed. I've dropped the scope parameter.

[agitter]

I have one more question that recurs across multiple files.