get_or_default()

[tower120]

Add

trait SparseHierarchyGetDefault: SparseHierarchy {
    /// This is faster then `unwrap_or(Default)`.
    fn data_or_default() -> Data<'a>
}

get_or_default() faster on full-sized-block array (zero-branching), and can be faster on CompactArray.

Operations that have idempotent-default resolve functions, can be SparseHierarchyGetDefault.
Idempotent-default function is f(Default state, Default state) = Default state.

[tower120]

Benchmarks shows that this will benefit only full-sized/non-compact SparseArray.
CompactSparseArray does not seems to have a technique that could get_or_deafult() faster then get().unwrap_or(&Default::default())

[tower120]

Implemented for SparseArray. But intersection, merge, etc does not utilize it yet.