perf: optimize aggregation with byte-level helpers and direct mmap

Introduce `byte_sum` and `byte_count_nonzero` to efficiently aggregate compact-int byte slices, bypassing per-element decoding and overflow map lookups. Refactor `sum()` and `count_nonzero()` across the matrix, reader, and traits modules to use direct memory-mapped slice iteration and idiomatic Rust iterators. Additionally, expose `MemoryIntIter` publicly and implement `IntoIterator` and `IntSlice` for `MemoryIntVec` to enable standard iteration and delegate aggregation to the new helpers.

src/obicompactvec/src/intmatrix.rs

@@ -11,7 +11,7 @@ use rayon::prelude::*;
 use crate::bitmatrix::{pairwise_matrix, pairwise2_matrix};
 use crate::builder::PersistentCompactIntVecBuilder;
 use crate::memoryintvec::MemoryIntVec;
-use crate::format::{HEADER_SIZE, OVERFLOW_ENTRY_SIZE, parse_index_entry, parse_overflow_entry};
+use crate::format::{byte_count_nonzero, byte_sum, HEADER_SIZE, OVERFLOW_ENTRY_SIZE, parse_index_entry, parse_overflow_entry};
 use crate::meta::MatrixMeta;
 use crate::reader::PersistentCompactIntVec;
 
@@ -230,16 +230,24 @@ impl PackedCompactIntMatrix {
 
     pub(crate) fn sum(&self) -> Array1<u64> {
         Array1::from_vec(
-            (0..self.n_cols).into_par_iter()
-                .map(|c| (0..self.n_rows).map(|s| self.get(c, s) as u64).sum())
+            self.columns.par_iter()
+                .map(|ci| {
+                    let primary = &self.mmap[ci.primary_start..ci.primary_start + self.n_rows];
+                    let overflow = (0..ci.n_overflow)
+                        .map(|i| parse_overflow_entry(&self.mmap, ci.data_offset, i).1);
+                    byte_sum(primary, overflow)
+                })
                 .collect()
         )
     }
 
     pub(crate) fn count_nonzero(&self) -> Array1<u64> {
         Array1::from_vec(
-            (0..self.n_cols).into_par_iter()
-                .map(|c| (0..self.n_rows).filter(|&s| self.get(c, s) > 0).count() as u64)
+            self.columns.par_iter()
+                .map(|ci| {
+                    let primary = &self.mmap[ci.primary_start..ci.primary_start + self.n_rows];
+                    byte_count_nonzero(primary)
+                })
                 .collect()
         )
     }