Add benchmarks for bulk-replace operations against Boost or STL

The C++ class implements a `replace_all` operation, that can be used to replace all occurrences of a substring or a character with a different string. The implementation is designed to minimize memory allocations and should be faster than `boost::algorithm::replace_all` in Boost and Python's `str.replace`. That intuition should be supported by evidence, so new benchmarks should be added.

Start by locating the C++ class and its replace_all operation, then identify the project's existing benchmark entry points. Add benchmarks covering substring and character replacement, compare the results with Boost's boost::algorithm::replace_all, STL operations, and Python's str.replace, and confirm that the results support the claimed performance advantage.

Written by the indexing model from the issue text.