Description

As I reported at elastic/elasticsearch#133989, I'd love to have a way to support multiple delimiters for the Path Hierarchy Tokenizer.

Currently, it only supports a single pattern for the delimiter parameter (default is /). This makes it difficult to tokenize both Windows (\\) and Linux (/) paths efficiently in the same index. Supporting multiple delimiters (such as both / and \\) would greatly improve usability for systems dealing with cross-platform file paths.

For example, a user may need to index file paths from both Windows and Linux environments and expects the analysis to work seamlessly regardless of path format. At the moment, the only workaround I found is to preprocess the data to normalize delimiters, which adds extra complexity.

Feature Request: Allow the path_hierarchy tokenizer to accept multiple delimiter patterns (e.g., an array of delimiters) so both / and \\ can be handled simultaneously.

Another possible implementation would be to create a new PathsHierarchyTokenizer (note the s) which implements this behavior.

Before working a such PR, I'd like to get your views about this proposal... May be I'm just wrong trying to do so.