Fixes benchmark / stress test numbers

[joelgallant]

You'll notice that here, they're using num_nanoseconds, which is different than you have it here.

subsec_nanos only returns, as you might guess, sub-seconds (only the fractional part).

This made me think that image captures were happening 100s of times faster than they actually were.

[pedrosland]

Hi Joel. Thanks for your contribution. I completely forgot that I was supposed to fix up that code.

I suppose that CI doesn't build the examples or something because this change doesn't compile for me. I need an as f64 to convert the u64 to f64. Does it compile for you? If not, do you want to fix that so I can merge?

Here's my compiler output.

error[E0277]: cannot add `f64` to `u64`
  --> examples/stress.rs:32:43
   |
32 |     let runtime_secs = duration.as_secs() + (duration.subsec_nanos() as f64 / 1_000_000_000.0);
   |                                           ^ no implementation for `u64 + f64`
   |
   = help: the trait `std::ops::Add<f64>` is not implemented for `u64`

[joelgallant]

Ah, oops. I wasn't cross-compiling it, so I figured the CI would catch any problems. I'll push what should be a fix (don't have a pi with me sorry).

[pedrosland]

Hi Joel. Thanks again. I have merged this manually and fixed CI to build the examples so we can catch errors 😄.