OK, this might be a stupid idea, but what about a simple old spirit level, a fancy oil based one that will never freeze, mounted to the side of the cockpit to give the pilots a emergency true level backup when all else fails?
It is about the only gravity based instrument on the aircraft (because ideally, you would want to keep a steady 1G straight towards the floor of the plane during a turn for passenger comfort).
All other instruments that relate to orientation are gyroscope based, so they constantly reference a point in infinite space. When you are being thrown about in turbulence, or a tight turn, gravity based instrumentation will be all over the place and give false readings.
Pretty much exactly why IFR (instrument flying) rated pilots are trained NOT to rely on the inner ear balance systems to know if they are flying straight and level etc. because the inner ear fluids can be easily tricked by standard flying manoeuvres into thinking you are in a screaming right hand dive when you are in a gentle left hand turn (if you don't have a visual reference to corroborate).
In a steady state, and for a given wing configuration (flaps/slats/spoiler positions), there's a causal relationship between angle of attack and speed (which also depends on air density, aircraft weight and load factor (how many gs you are pulling)), such that the indicated airspeed (which measures ram air pressure in a pitot tube, and fortuitously cancels out the density dependence) has historically been the proxy for angle of attack - and, at least in small airplanes, it still is.
Even small airplanes often have a separate stall warning device, which is triggered by the movement of the leading-edge stagnation point (where the flow divides between going above and below the wing) as the angle of attack changes. In its simplest form, it is operated by a small tab which is flipped up when the stagnation point moves below it.
In the case of the Lion Air crash, the problem was not that the crew were unaware of the airplane's attitude (it was daylight in good weather, and they also had functioning airspeed indicators), it was that the control system kept resetting the longitudinal trim incorrectly in response to a faulty angle-of-attack sensor, and the crew did not follow the procedure to disable it from doing so.