Add experimental MLX backend and engine with imagegen support (#13648)

* WIP - MLX backend with gemma3

* MLX: add cmake and go tag build toggles

To build the new MLX backend code:
  cmake --preset MLX
  cmake --build --preset MLX --parallel
  cmake --install build --component MLX
  go build -tags mlx .

Note: the main.go entrypoint for the MLX engine will change in a follow up commit.

* add experimental image generation runtime

* add experimental image generation runtime

* MLX: wire up cuda build for linux

* MLX: get dependencies correct and dedup

This is still too large for a unified github artifact, but is now "correct" for the mlx_cuda_v13
directory.

* fix relative link bug in dedup

* Add darwin build and readme

* add go build tag for mlx dependent code and wire up build_darwin.sh

* lint cleanup

* macos: build mlx for x86

This will be CPU only.

* cuda build instructions and fix drift from mlx bump

* stale comment

* Delete agent helper doc

* Clean up readme.md

* Revise README for tokenizer clarity and details

Updated README to clarify tokenizer functionality and removed correctness section.

---------
Co-authored-by: jmorganca <jmorganca@gmail.com>

x/imagegen/tokenizer/testdata/mini_llama.json

+{"model": {"type": "BPE", "vocab": {"!": 0, "\"": 1, "#": 2, "$": 3, "%": 4, "&": 5, "'": 6, "(": 7, ")": 8, "*": 9, "+": 10, ",": 11, "-": 12, ".": 13, "/": 14, "0": 15, "1": 16, "2": 17, "3": 18, "4": 19, "5": 20, "6": 21, "7": 22, "8": 23, "9": 24, ":": 25, ";": 26, "<": 27, "=": 28, ">": 29, "?": 30, "@": 31, "A": 32, "B": 33, "C": 34, "D": 35, "E": 36, "F": 37, "G": 38, "H": 39, "I": 40, "J": 41, "K": 42, "L": 43, "M": 44, "N": 45, "O": 46, "P": 47, "Q": 48, "R": 49, "S": 50, "T": 51, "U": 52, "V": 53, "fé": 59958, "W": 54, "X": 55, "Y": 56, "Z": 57, "[": 58, "\\": 59, "]": 60, "^": 61, "_": 62, "`": 63, "a": 64, "b": 65, "c": 66, "d": 67, "e": 68, "f": 69, "g": 70, "h": 71, "i": 72, "j": 73, "k": 74, "l": 75, "m": 76, "n": 77, "o": 78, "p": 79, "r": 81, "q": 80, "s": 82, "t": 83, "u": 84, "v": 85, "w": 86, "x": 87, "y": 88, "z": 89, "{": 90, "|": 91, "}": 92, "~": 93, "¡": 94, "¢": 95, "£": 96, "¤": 97, "¥": 98, "¦": 99, "§": 100, "¨": 101, "World": 10343, "©": 102, "ª": 103, "«": 104, "¬": 105, "®": 106, "world": 14957, "¯": 107, "°": 108, "±": 109, "²": 110, "³": 111, "´": 112, "µ": 113, "¶": 114, "·": 115, "¸": 116, "¹": 117, "º": 118, "»": 119, "¼": 120, "½": 121, "¾": 122, "¿": 123, "À": 124, "Á": 125, "Â": 126, "Ã": 127, "Ä": 128, "Å": 129, "Æ": 130, "Ç": 131, "È": 132, "É": 133, "Ê": 134, "Ë": 135, "Ì": 136, "Í": 137, "Î": 138, "Ï": 139, "Ð": 140, "Ñ": 141, "Ò": 142, "Ó": 143, "Ô": 144, "Õ": 145, "Ö": 146, "×": 147, "Ø": 148, "Ù": 149, "Ú": 150, "Û": 151, "Ü": 152, "Ý": 153, "Þ": 154, "ß": 155, "à": 156, "á": 157, "â": 158, "ã": 159, "ä": 160, "å": 161, "æ": 162, "ç": 163, "è": 164, "é": 165, "ê": 166, "ë": 167, "ì": 168, "Ġhello": 24748, "í": 169, "î": 170, "ï": 171, "ð": 172, "ñ": 173, "Hello": 9906, "ò": 174, "ó": 175, "ô": 176, "õ": 177, "ö": 178, "Ġ{}": 4792, "÷": 179, "ø": 180, "ù": 181, "ú": 182, "û": 183, "ü": 184, "ý": 185, "þ": 186, "ÿ": 187, "Ā": 188, "ā": 189, "Ă": 190, "ă": 191, "Ċ": 198, "Ą": 192, "ą": 193, "Ć": 194, "ć": 195, "Ĉ": 196, "ĉ": 197, "ċ": 199, "Č": 200, "č": 201, "Ď": 202, "ď": 203, "Đ": 204, "đ": 205, "Ē": 206, "ē": 207, "Ĕ": 208, "ĕ": 209, "Ė": 210, "ė": 211, "Ę": 212, "ę": 213, "Ġ": 220, "Ě": 214, "ě": 215, "Ĝ": 216, "ĝ": 217, "Ğ": 218, "ğ": 219, "ġ": 221, "Ģ": 222, "ģ": 223, "Ĥ": 224, "ĥ": 225, "Ħ": 226, "ħ": 227, "Ĩ": 228, "ĩ": 229, "Ī": 230, "ī": 231, "Ĭ": 232, "ĭ": 233, "Į": 234, "į": 235, "İ": 236, "ı": 237, "IJ": 238, "ij": 239, "Ĵ": 240, "ĵ": 241, "Ķ": 242, "ķ": 243, "ĸ": 244, "Ĺ": 245, "ĺ": 246, "Ļ": 247, "ļ": 248, "Ľ": 249, "ĠĠ": 256, "ľ": 250, "Ŀ": 251, "ŀ": 252, "Ł": 253, "rer": 38149, "ĠĠĠ": 262, "ł": 254, "Ń": 255, "'m": 2846, "'re": 2351, "can": 4919, "func": 2900, "()": 368, "Ġworld": 1917, "Ġmain": 1925, "00": 410, "123": 4513, "000": 931, "ca": 936, "'t": 956, "é": 978, "hello": 15339, "Ġw": 289, "orld": 1410, "Ġwor": 4191, "ld": 509, "main": 3902, "Ġm": 296, "ain": 467, "Ġma": 7643, "in": 258, "Ġmai": 17154, "re": 265, "'r": 97670, "unc": 1371, "fun": 12158, "fu": 33721, "nc": 1031, "ma": 1764, "mai": 77585, "wor": 50810, "or": 269, "Ġwo": 24670, "23": 1419, "12": 717, "{}": 6390, "Ġ{": 314, "an": 276, "ello": 4896, "Hel": 33813, "lo": 385, "Hell": 81394, "un": 359, "hel": 50222, "hell": 57195, "ai": 2192, "wo": 1146, "Ġh": 305, "Ġhel": 11591, "Ġhell": 15123, "el": 301, "He": 1548, "er": 261, "he": 383, "ell": 616, "ll": 657}, "merges": ["Ġ Ġ", "Ġ ĠĠ", "ĠĠ Ġ", "( )", "0 0", "0 00", "00 0", "c a", "' t", "à ©", "Ġ world", "Ġw orld", "Ġwor ld", "Ġ main", "Ġm ain", "Ġma in", "Ġmai n", "' re", "'r e", "' m", "f unc", "fun c", "fu nc", "m ain", "ma in", "mai n", "Ġ wor", "Ġw or", "Ġwo r", "1 23", "12 3", "Ġ {}", "Ġ{ }", "c an", "ca n", "{ }", "Ġ ma", "Ġm a", "H ello", "Hel lo", "Hell o", "W orld", "f un", "fu n", "w orld", "wor ld", "h ello", "hel lo", "hell o", "Ġ mai", "Ġm ai", "Ġma i", "Ġ wo", "Ġw o", "Ġ hello", "Ġh ello", "Ġhel lo", "Ġhell o", "f u", "H el", "He l", "r er", "re r", "h el", "he l", "w or", "wo r", "h ell", "he ll", "hel l", "f é", "m ai", "ma i", "H ell", "He ll", "Hel l", "' r"]}, "pre_tokenizer": {"type": "Sequence", "pretokenizers": [{"type": "Split", "pattern": {"Regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}{1,3}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+"}, "behavior": "Isolated", "invert": false}, {"type": "ByteLevel", "add_prefix_space": false, "trim_offsets": true, "use_regex": false}]}, "decoder": {"type": "ByteLevel", "add_prefix_space": true, "trim_offsets": true, "use_regex": true}, "added_tokens": [{"id": 128000, "content": "<|begin_of_text|>", "special": true}, {"id": 128001, "content": "<|end_of_text|>", "special": true}]}
\ No newline at end of file

x/ml/backend/backend.go

+package backend
+
+// _ "github.com/ollama/ollama/x/ml/backend/mlx"