fix: flatten nested arrays in (co)lexicographic order

This commit addresses a bug in flattening behavior. Prior to this
commit, if provided a nested array, the implementation would flatten
in lexicographic order (i.e., innermost nested array dimensions
iterate fastest). However, this lead to surprising behavior when the
desired storage order was column-major in which an ndarray did not
match the visual layout of the input nested arrays. This commit
fixes this issue and matches NumPy behavior in which nested arrays
are flattened such that the storage order matches the desired output
array storage order.

Ref: #268

Author: kgryte

Diff for: lib/node_modules/@stdlib/ndarray/array/lib/main.js

@@ -321,7 +321,7 @@ function array() {
 		}
 	} else if ( buffer ) {
 		if ( btype === 'generic' && opts.flatten ) {
-			buffer = flatten( buffer, osh || arrayShape( buffer ), false );
+			buffer = flatten( buffer, osh || arrayShape( buffer ), order === 'column-major' );
 		}
 		if ( buffer.length !== len ) {
 			throw new RangeError( 'invalid arguments. Array shape is incompatible with provided data source. Number of data source elements does not match array shape.' );