Reworked the set_suffix_array_mem method of SufrFile (https://github.com/TravisWheelerLab/sufr/blob/main/libsufr/src/sufr_file.rs#L514C8-L514C28) to avoid redundant loading of the suffix array into memory:

1. If the query's max_query_len is None, it defaults to self.text_len.to_usize(), which will be equal to built_max_query_len if the suffix array was also built with max_query_len set to None. This means that every time set_suffix_array_mem is called, the condition on line 539 is triggered, leading to the entire suffix array being read from disk. If multiple queries are made with the Options max_query_len: None, the suffix array will be read every single time.
2. If the query len is different from the built query len, there is a condition for early-return if the loaded suffix array is nonempty and its internal max query len self.suffix_array_mem_mql matches the one from the Options. However, self.suffix_array_mem_mql is None by default and only gets set when the suffix array gets read from a non-cached file.
3. Afaict, in every circumstance, by the end of set_suffix_array_mem, the suffix array in memory will have a max query len matching the Options max query len, so the change I made sets self.suffix_array_mem_mql = Some(max_query_len); whenever the function does not return early.
4. To avoid the reloading issue in 1., I also moved the early-return in front of the other conditions, so that whenever a suffix array has been loaded into memory, and its max query length matches that of the Options' max query length, the method returns without doing redundant work.