[recipe] Add recipe for Decomposed GRU

[BalyshevArtem]

This pr adds recipe for decomposed GRU.

from #13602

ONE-DCO-1.0-Signed-off-by: Artem Balyshev [email protected]

[BalyshevArtem]

For generation I used the following python script to get tflite model

import tensorflow as tf
from tensorflow import keras
import numpy as np

activation = 'tanh'
model = tf.keras.models.Sequential([
    tf.keras.Input(shape=(1,2)),
    tf.keras.layers.GRU(units=1, activation=activation, return_sequences=False, time_major=True),
])

model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.001))

model.summary()

run_model = tf.function(lambda x: model(x))

# This is important, let's fix the input size.
BATCH_SIZE = 1
X = 1
Y = 2
concrete_func = run_model.get_concrete_function(
    tf.TensorSpec([BATCH_SIZE, X,Y], model.inputs[0].dtype))

# model directory.
MODEL_DIR = "keras_model"
model.save(MODEL_DIR, save_format="tf", signatures=concrete_func)

converter = tf.lite.TFLiteConverter.from_saved_model(MODEL_DIR)
converter.experimental_new_converter = True
converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS,
                                       ]
converted_model = converter.convert()
save_to = "my_gru.tflite"
if save_to is not None:
    with open(save_to, 'wb') as tf_lite_file:
        tf_lite_file.write(converted_model)
                                          

To obtain result recipe I used ./tflchef-reverse

[seanshpark]

#13602 failed with CI pass. I need that OK.
there is no dredd test. plz add test.rule file too.

[SlavikMIPT]

LGTM

[seanshpark]

#13649 (comment)