Fix ffmpeg info parser to avoid unexpected metadata lose and type error in certain case.

[toyaji]

This is related to [#2311] and [#1860].

At first, I got a TypeError from flowing line for self._last_metadata_field_added was "displaymatrix" which is casted integer, then trying operate int + str. I can easily avoid this TypeError by type conversion or check if the last metadata field was catsted. But fundamentally, why this code try to concat the fields that are not multiple line metadata?

moviepy/moviepy/video/io/ffmpeg_reader.py

Lines 545 to 548 in 4dd5fe4

	# multiline metadata value parsing
	if field == "":
	field = self._last_metadata_field_added
	value = self._current_stream["metadata"][field] + "\n" + value

Then I found [#2311] caused this issue because changed code ignore none "key:value" metadata and return empty string "" as default. So above line is triggered which was intended to multiple line metadata concat.

So I think we need to handle none "key:value" metadata rather than not just ignoring them. Following keys are none "key:value" metadata I've just found in dump.c in ffmpeg original repo. And I want add them on parse_metadata_field_value() to parse properly.

• Mastering Display Metadata,
• Ambient Viewing Environment,
• Content Light Level Metadata,

Additionally, I found for video stream there is sidedata besides metadata but currently parsed as metadata. So I separate sidedata from metadata for video stream.