Feature: Implement Compute Functions for FixedSizeListArray

[connortsui20]

Tracking Issue: #4372

Implements the same compute functions as ListArray, but for FixedSizeListArray.

• Take: Select elements by indices with optimized paths for nullable/non-nullable data
• Filter: Apply boolean masks with density-based expansion strategy (not sure it this is a premature optimization)
• Cast
• IsConstant: Check if all lists in the array are identical (this is super expensive and it might be worth revisiting the compute function characteristics itself, cc @gatesn)
• Mask
• Fuzzing: Extended arbitrary data generation to support FixedSizeListArray
• Arrow conversion
• Refactored ListArray filter compute function with a similar density check

The two complicated compute functions here are filter and take, with take being significantly more complicated. Those are accompanied by tests since the logic is a lot harder.

Nice-to-have-but-I'm-too-tired:

• Tests for each compute function (missing for cast, is_constant, and mask)
• Refactor the ListArray take compute function

[codecov]

Codecov Report

❌ Patch coverage is 88.47385% with 108 lines in your changes missing coverage. Please review.
✅ Project coverage is 88.80%. Comparing base (4caa917) to head (ad56a75).
⚠️ Report is 23 commits behind head on develop.

Files with missing lines	Patch %	Lines
...rtex-array/src/arrow/compute/to_arrow/canonical.rs	6.06%	31 Missing ⚠️
...x-array/src/arrays/fixed_size_list/compute/cast.rs	0.00%	22 Missing ⚠️
.../src/arrays/fixed_size_list/compute/is_constant.rs	0.00%	16 Missing ⚠️
...x-array/src/arrays/fixed_size_list/compute/mask.rs	0.00%	11 Missing ⚠️
encodings/runend/src/compute/filter.rs	68.18%	7 Missing ⚠️
vortex-array/src/arrays/list/compute/filter.rs	92.06%	5 Missing ⚠️
vortex-dtype/src/dtype.rs	54.54%	5 Missing ⚠️
...ortex-array/src/compute/conformance/consistency.rs	0.00%	4 Missing ⚠️
...array/src/arrays/fixed_size_list/compute/filter.rs	94.23%	3 Missing ⚠️
...x-array/src/arrays/fixed_size_list/compute/take.rs	98.03%	2 Missing ⚠️
... and 2 more

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[robert3005]

I think we can simplify things a little bit

[connortsui20]

This seems somewhat error-prone, should I create a as_fixed_size_list_element_opt that only returns the type if it is DType::FixedSizeList and have as_list_element_opt return None in that case?

[robert3005]

I don't think s we should maybe just make this work?

[robert3005]

Having thought about this more I think you want to separate this two out. If FixedSizeList is a dtype then it's elements are distinct from the list elements

[connortsui20]

I separated them out:

    /// Get the inner element dtype if `self` is a [`DType::List`], otherwise returns `None`.
    ///
    /// Note that this does _not_ return `Some` if `self` is a [`DType::FixedSizeList`].
    pub fn as_list_element_opt(&self) -> Option<&Arc<DType>> {
        if let List(edt, _) = self {
            Some(edt)
        } else {
            None
        }
    }

    /// Get the inner element dtype if `self` is a [`DType::FixedSizeList`], otherwise returns
    /// `None`.
    ///
    /// Note that this does _not_ return `Some` if `self` is a [`DType::List`].
    pub fn as_fixed_size_list_element_opt(&self) -> Option<&Arc<DType>> {
        if let FixedSizeList(edt, ..) = self {
            Some(edt)
        } else {
            None
        }
    }

    /// Get the inner element dtype if `self` is **either** a [`DType::List`] or a
    /// [`DType::FixedSizeList`], otherwise returns `None`
    pub fn as_any_size_list_element_opt(&self) -> Option<&Arc<DType>> {
        if let FixedSizeList(edt, ..) = self {
            Some(edt)
        } else if let List(edt, ..) = self {
            Some(edt)
        } else {
            None
        }
    }

[connortsui20]

Yeah this is pretty bad. I'm also realizing that this will also check the dtypes are equal every single time. Maybe a eq_value or something is needed? I'll probably leave this for later.

[robert3005]

Only have some nits