The STM32F071V8T7TR: The 87 I/O pins in a 100-LQFP package give you room for parallel interfaces and plenty of GPIO, though the 14x14 mm body means the board area is a fixed cost to budget for.
The 48 MHz core clock is the performance ceiling for this part. It's a Cortex-M0, so there's no hardware multiply-accumulate or FPU — the integer ALU handles bit-banging, state machines, and simple PID loops cleanly, but you'll feel the limit if you try software filtering above a few kilohertz. The 16 KB RAM is the real bottleneck: it fits a modest RTOS heap plus a few buffers, but anything needing large frame buffers or deep look-up tables will push you to the STM32F0 or F3 family.
The 100-LQFP package is a 14x14 mm body with 0.5 mm pitch leads. It's hand-reworkable with hot air if you're careful — the leads are exposed, so you can flux and drag-solder without lifting pads.