Use per-attribute node.attribute() for cell t/s/r

Follow-up to the rows_generator optimization. Fetch the three cell
attributes with individual node.attribute('t'/'s'/'r') C calls instead of
building a per-cell attribute_hash and indexing it. With hundreds of
thousands of cells the per-cell Hash allocation dominates, so three cheap
lookups are both faster and leaner.

Same 20,000-row x 40-col benchmark, vs the attribute_hash version:

  | Cell-attr approach     | Median  | Allocations |
  |------------------------|---------|-------------|
  | node.attribute_hash    | ~1.10s  | 9,226,711   |
  | node.attribute() x3    | ~1.08s  | 7,626,711   |

  => ~3-4% faster, 1.6M (~17%) fewer allocations (deterministic: ~640K
     cells x the per-cell hash no longer built).

Output byte-identical across rows, simple_rows, and rows_with_meta_data on
both default- and prefixed-namespace files. Suite green (47 examples).
The row-opener node keeps attribute_hash (only ~one per row, and it
preserves the exact meta-data row hash).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: connorshea

lib/creek/sheet.rb

@@ -131,12 +131,13 @@ def rows_generator(include_meta_data = false, use_simple_rows_format = false)
                 cells[cell] = convert(node.value, cell_type, cell_style_idx)
               end
             elsif node_name == name_c && node_type == opener
-              # attribute_hash avoids the namespaces lookup + merge that
-              # Reader#attributes performs on every call; we only need t/s/r.
-              attributes     = node.attribute_hash
-              cell_type      = attributes['t']
-              cell_style_idx = attributes['s']
-              cell           = attributes['r']
+              # Fetch the three attributes individually rather than via
+              # attribute_hash/attributes: with hundreds of thousands of cells
+              # the per-cell Hash allocation dominates, so three cheap C lookups
+              # are both faster and leaner than building and indexing a hash.
+              cell_type      = node.attribute('t')
+              cell_style_idx = node.attribute('s')
+              cell           = node.attribute('r')
             elsif node_name == name_row && node_type == opener
               row = node.attribute_hash
               row['cells'] = {}