fix(spectroscope): zero-pad fft samples to expected size

[dotacow]

First of all, thanks for making this project, it earned a place in my terminal, it is very neat & nice to look at, I also want to learn more about DSP and this is a great output of doing that.

fixes a panic that occurs when not using the PulseAudio backend

I initially ran the program

./target/debug/scope-tui audio # I also tried matching buffer size with -b512,2048, etc.,  didn't seem to help, expected size just kept changing to passed size/2

then when I tabbed my way to the spectroscope view:

o/registry/src/index.crates.io-1949cf8c6b5b557f/rustfft-6.3.0/src/common.rs:19:5:
Provided FFT buffer was too small. Expected len = 2048, got len = 512
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

I recreated the panic with gdb running, found the culprit at src/display/spectroscope.rs:163, so I just prefixed it with a check that zero-pads the tmp buffer to match the expected sample size.

I explain my process because I don't know much about the Linux audio stack, so I don't know why the accumulated buffer doesn't match the expected size, but I suspect it may be because PulseAudio protects us from this failure by automatically padding the audio peripheral queue to reach 2048 samples.