create pre-tokenizer with surface-projection does not override dictionary-projection

[mh-northlander]

Creating pre-tokenizer with surface-projection specified does not overrides the projection of dictionary.

In test:

sudachi.rs/python/tests/test_pretokenizers.py

Lines 105 to 119 in 232d9ee

	def test_projection_surface_override(self):
	dictobj = sudachipy.Dictionary(config=sudachipy.config.Config(projection="reading"))
	pretok = dictobj.pre_tokenizer(sudachipy.SplitMode.A, projection="surface")
	vocab = {
	"[UNK]": 0,
	"サケ": 1,
	"ヒト": 2,
	"ノム": 3,
	"ヲ": 5,
	"外国人参政権": 4
	}
	tok = tokenizers.Tokenizer(WordLevel(vocab, unk_token="[UNK]"))
	tok.pre_tokenizer = pretok
	res = tok.encode("酒を飲む人")
	self.assertEqual(res.ids, [1, 5, 3, 2])

Is this intentional or bug?