You can either use a high-speed stepper motor, the kind used in hard disks and CD-ROM drives, or even a brushless DC motor. While you may order or otherwise improvise the flywheel itself, you won't get a better precision than using a stack of hard disk plates; these provide precision in excess of anything you can have custom-made on reasonable budget and provide a decent density; mass easily regulated through changing the number of plates used.
The wheel speed readout, while helpful, is not essential; in case of brushless DC it's achievable from the same Hall sensor that drives the motor; in case of steppers you just assume the motor follows your control and only monitor current to determine if it's not falling behind or blocked. But these, again, are non-critical. You need to be able to regulate the speed, but not its precise value, only delta, "faster/slower than now".
You need rotational speed of the satellite, and use that to drive the wheel through a closed-loop algorithm (PID or similar). Accelerometers are helpful in estimating that - after substracting common bias of Earth's gravity (not applicable in space but your major headache when testing) you will have the remaining centrifugal acceleration value. Estabilishing direction of rotation will be a bit more tricky. Likely accelerometer response to changing speed of the reaction wheel will tell you - if you accelerate your wheel clockwise (and craft counterclockwise) increasing centrifugal acceleration means it was spinning clockwise in the first place; decreasing - that you just reduced its angular velocity.
Once you get the spin in check and reduced to reasonable values (a few rotations per second max) then you may employ a gyroscope to arrest the rotation entirely. (it will be saturated before that, producing no usable output).
Now for control: Whether using DC motor, or stepper, you'll need a microcontroller, and a driver chip. Microcontroller to read the sensors, provide the calculations and control signal, the driver to convert signal into actual driving current. In case of DC, that will be an amplifier, capable of inverting the signal, or a H-bridge, where you drive with PWM (as opposed to voltage from DAC for use with amplifier) and it enables you to switch the direction of the motor; steppers use specialized stepper motor driver chips. The reason is that your microcontroller will be able to produce maybe 50 milliampers of current on its output pins, and your motor needs way more to run; also in case of steppers it's an awful bother to control the sequence of coils activation; the driver lets you output "direction" and "step" instead of requiring which coils need to be on and which ones off.
Your control program will be a major piece of work. You'll need sensor readout, conversion to useful units, control algorithm, and output, plus external "steering" input so that your reaction wheel can rotate the satellite to desired angle, or perform desaturation operation, instead of just holding the platform immobile in one position forever. You'll need data sheets of the components you use, and a lot of stuff to learn. Not a lesson for a single StackeExchange answer.
Last, the pointing accuracy. Either you derive it by adding expectable errors, or determine it experimentally. Hang your rig on a thin string (fishing line), so it can spin freely and be stabilized in the spin axis by the reaction wheel. Attach a small mirror, shine a laser at that mirror and determine how stable you can keep the dot of the reflected ray. Using trigonometry and determining how much the dot travels while stabilized after angle change or external disturbance, you can determine how accurately you control the angle.