feat(image): add custom RLE decoder (#4332)

Signed-off-by: Xu Xingliang <xuxingliang@xiaomi.com>

docs/libs/index.rst

@@ -19,3 +19,4 @@
     barcode
     rlottie
     ffmpeg
+    rle

docs/libs/rle.rst

@@ -0,0 +1,84 @@
+RLE Decoder
+===========
+
+The RLE Decoder is a part of LVGL that is responsible for decoding RLE
+compressed original LVGL binary files. The file name suffix is always set
+to '.rle' and the file header is added with another 64-bit value that
+includes a 32-bit magic number and 32-bit RLE information.
+
+The decoder supports both variable and file as image sources. The original
+binary data is directly decompressed to RAM, and further decoded by LVGL's
+built-in decoder.
+
+Benefits
+--------
+
+Based on test result from a watch project. Most of the images can be compressed
+to save more than 70% space as show in below statistic. It shows the file count
+of every compress level. For rare conditions, RLE compress may increase the file
+size if there's no large repetition in data.
+
+.. image:: /libs/rle-compress-statistics.png
+   :alt: RLE compress statistics from a watch project
+   :align: center
+
+
+Theory
+------
+
+The RLE algorithm is a simple compression algorithm that is based on the fact that
+the for many pixels, the color is the same. The algorithm simply counts how many
+repeated data are there and store the count value and the color value.
+If the coming pixels are not repeated, it stores the non-repeat count value and
+original color value. For more details, the script used to compress the image
+can be found from `lvgl/script/LVGLImage.py`.
+
+.. code:: python
+
+    def rle_compress(self, data: bytearray, blksize: int, threshold=16):
+        index = 0
+        data_len = len(data)
+        compressed_data = []
+        while index < data_len:
+            memview = memoryview(data)
+            repeat_cnt = self.get_repeat_count(
+                memview[index:], blksize)
+            if repeat_cnt == 0:
+                # done
+                break
+            elif repeat_cnt < threshold:
+                nonrepeat_cnt = self.get_nonrepeat_count(
+                    memview[index:], blksize, threshold)
+                ctrl_byte = uint8_t(nonrepeat_cnt | 0x80)
+                compressed_data.append(ctrl_byte)
+                compressed_data.append(
+                    memview[index: index + nonrepeat_cnt*blksize])
+                index += nonrepeat_cnt * blksize
+            else:
+                ctrl_byte = uint8_t(repeat_cnt)
+                compressed_data.append(ctrl_byte)
+                compressed_data.append(memview[index: index + blksize])
+                index += repeat_cnt * blksize
+
+        return b"".join(compressed_data)
+
+
+Usage
+-----
+
+To use the RLE Decoder, ensure that `LV_USE_RLE` is defined and set to `1`.
+The RLE image can be used same as other images.
+
+.. code:: c
+
+   lv_image_set_src(img, "path/to/image.rle");
+
+Generate RLE images
+-------------------
+
+The image can be directly generated using script `lvgl/script/LVGLImage.py`
+
+
+.. code:: bash
+
+   ./script/LVGLImage.py --ofmt RLE --cf I8 cogwheel.png