Use of key_prefix in _make_cache_key_query_string()

[willcroft]

In looking at the implementation of @cache.cached(query_string=True), I note that key_prefix is completely ignored. Is there an intentional design decision behind this that would otherwise cause issues?

For example, the simplest version of this change might be:

def _make_cache_key_query_string():
    ...

    if isinstance(key_prefix, str) and "%s" in key_prefix:
        cache_key = (key_prefix % request.path) + cache_hash
    else:
        cache_key = request.path + cache_hash

    return cache_key

(Here I only provide the corresponding code path for_make_cache_key's case when use_request=True since setting query_string=True implicitly requires the request.)

Additionally, might it be useful to have a separator character between the prefix/path and the cache_hash, perhaps :? This might aid future work in invalidating cache keys for all query string variants of the same request path.

Very happy to provide a PR if either are deemed useful additions!