videotestsrc: fix race condition when clearing cached buffer

Discovered a crash due to race condition with src->cached. It is non-null when checked on L1307, but gets cleared from a separate thread on L767 before the null pointer is then accessed on L1321.

My solution is to use a flag to invalidate the cache rather than clear it directly.

edit: after discussion w/@slomo, I added a mutex around cached buffer.

I believe the bug first appears in gstreamer 1.22

python script to reproduce:

import sys
import traceback
import threading
import time
import gi
gi.require_version('Gst', '1.0')
from gi.repository import Gst, GObject, GLib

Gst.init(sys.argv)

ll = 'videotestsrc name=tstsrc is-live=true pattern=17 ! queue ! fakesink'
def on_message(bus: Gst.Bus, message: Gst.Message, loop: GLib.MainLoop):
    mtype = message.type
    if mtype == Gst.MessageType.EOS:
        print("End of stream")
        loop.quit()

    elif mtype == Gst.MessageType.ERROR:
        err, debug = message.parse_error()
        print(err, debug)
        loop.quit()

    elif mtype == Gst.MessageType.WARNING:
        err, debug = message.parse_warning()
        print(err, debug)

    return True

pipeline = Gst.parse_launch(ll)
bus = pipeline.get_bus()
bus.add_signal_watch()

pipeline.set_state(Gst.State.PLAYING)
print('started pipeline')

loop = GLib.MainLoop()

bus.connect("message", on_message, loop)

def hammer():
    global pipeline
    while True:
        ele = pipeline.get_by_name('tstsrc')
        time.sleep(.0001)
        ele.set_property('foreground-color', int(0x1234567))

time.sleep(5)
t = threading.Thread(target=hammer)
t.start()

try:
    loop.run()
except Exception:
    traceback.print_exc()
    loop.quit()

pipeline.set_state(Gst.State.NULL)