Revise baudrate calculations

[attie-argentum]

As discussed in #40, the baudrate calculations should not consider SJW as part of the bit time.

This PR revises the timing calculations to exclude SJW, performance is also improved while maintaining a sample point between 75-95%.

Out-of-bounds SJW values are clamped, and the value is kept as high as possible for improved interoperability.
The lookup of pre-calculated timing configuration is also improved.

---

During calculation, all prescaler values that are unable to achieve the desired baudrate are immediately skipped, and restricting BS1/BS2 values as below allows us to also discard a large number of iterations that would produce inferior configurations.

Compared to the previous code, we gain 15x potential configuration options per prescaler value, and also discard 3x inferior options (early sample point)... this is 45x total vs 33x previously.

The number of time quanta for each BS1 / BS2 value is shown below:

This translates into a sample point as shown in this table - all values <75% are ignored (previous options ringed in blue).

This means that we can limit BS2 to values between 1 and 5, and BS1 may then start from (BS2 * 3) - 1.

[pazi88]

Sorry for delay. This looks very good. I will test this little bit first.

[attie-argentum]

No problem!

We've actually had further issues related to small BS1 and/or BS2 values (limiting the efficacy of SJW) and remote nodes starting their ACKs too early (causing an Error Frames)... I think it will be quite difficult to build this into a generic "calculate the parameters" function at runtime, so we have opted for a patch with manually selected timing configuration that works for our system... I wonder if it might be useful to consider a formal API for setting custom timings as well?