Feature Request: Support `05Up` RoundingMode

[LambdaAlpha]

According to wikipedia, there are many rounding modes which are not included in RoundingMode. There is a special rounding mode which is used to avoid getting a potentially wrong result after multiple roundings. Python has provided this rounding mode in its decimal library and called it ROUND_05UP. I request adding 05Up to RoundingMode, because this mode enables users to implement their custom rounding strategies, by rounding result with 05Up mode and with precision target_precision + 1 and then doing a custom rounding again to target_precision.

The RoundingMode has a round_pair method

fn round_pair(&self, sign: Sign, (d0, d1): (u8, u8), trailing_zeros: bool) -> u8;

The 05Up mode implementation of this method is like

fn round_pair(sign: Sign, (d0, d1): (u8, u8), trailing_zeros: bool) -> u8 {
    if d0 != 0 && d0 != 5 {
        return d0;
    }
    if d1 == 0 && trailing_zeros {
        return d0;
    }
    return d0 + 1;
}

[akubera]

Can you write some examples I can throw in the unit tests for me?

[akubera]

(this is a note mostly for myself)

Output of Python script rounding decimals to 1 digit of precision:

0.00 -> 0
0.01 -> 1
0.02 -> 1
1.01 -> 1
4.01 -> 4
5.00 -> 5
5.01 -> 6
6.01 -> 6

more examples:

-0.999 -> -0.9
-0.900 -> -0.9
-0.899 -> -0.8
-0.800 -> -0.8
-0.799 -> -0.7
-0.700 -> -0.7
-0.699 -> -0.6
-0.600 -> -0.6
-0.599 -> -0.6    ## Not -0.5
-0.500 -> -0.5
-0.499 -> -0.4
-0.400 -> -0.4
-0.399 -> -0.3
-0.300 -> -0.3
-0.299 -> -0.2
-0.200 -> -0.2
-0.199 -> -0.1
-0.100 -> -0.1
-0.099 -> -0.09
0.000 -> 0
0.001 -> 0.001
0.100 -> 0.1
0.101 -> 0.1
0.200 -> 0.2
0.201 -> 0.2
0.300 -> 0.3
0.301 -> 0.3
0.400 -> 0.4
0.401 -> 0.4
0.500 -> 0.5
0.501 -> 0.6
0.600 -> 0.6
0.601 -> 0.6
0.700 -> 0.7
0.701 -> 0.7
0.800 -> 0.8
0.801 -> 0.8
0.900 -> 0.9
0.901 -> 0.9
1.000 -> 1
1.001 -> 1
1.100 -> 1
1.101 -> 1
1.200 -> 1
1.201 -> 1
1.300 -> 1
1.301 -> 1
1.400 -> 1
1.401 -> 1
1.500 -> 1
1.501 -> 1
1.600 -> 1
1.601 -> 1
1.700 -> 1
1.701 -> 1
1.800 -> 1
1.801 -> 1
1.900 -> 1
1.901 -> 1
2.000 -> 2
2.001 -> 2
2.100 -> 2
2.101 -> 2
2.200 -> 2
...
4.700 -> 4
4.701 -> 4
4.800 -> 4
4.801 -> 4
4.900 -> 4
4.901 -> 4
5.000 -> 5   # only '5' result is rounding 5.000
5.001 -> 6
5.100 -> 6
5.101 -> 6
5.200 -> 6
5.201 -> 6
5.300 -> 6
5.301 -> 6
5.400 -> 6
5.401 -> 6
5.500 -> 6
5.501 -> 6
5.600 -> 6
5.601 -> 6
5.700 -> 6
5.701 -> 6
5.800 -> 6
5.801 -> 6
5.900 -> 6
5.901 -> 6
6.000 -> 6
6.001 -> 6
6.100 -> 6
6.101 -> 6
6.200 -> 6
6.201 -> 6
6.300 -> 6
6.301 -> 6
6.400 -> 6
6.401 -> 6
6.500 -> 6
6.501 -> 6
6.600 -> 6
6.601 -> 6
6.700 -> 6
6.701 -> 6
6.800 -> 6
6.801 -> 6
6.900 -> 6
6.901 -> 6

[akubera]

Ok, I think this is all we need: 92b0ad8 (besides tests)

[LambdaAlpha]

Examples

9.99x →  9.0
9.6  →  9.0
6.9 →  6.0
5.1 →  6.0
5.00x  →  6.0
5.0 →  5.0
4.99x →  4.0
4.6  →  4.0
1.6  →  1.0
0.1  →  1.0
0.00x  →  1.0
0.0 →  0.0
-0.00x  →  -1.0
-0.1  →  -1.0
-1.6  →  -1.0
-4.6  →  -4.0
-4.99x →  -4.0
-5.0 →  -5.0
-5.00x  →  -6.0
-5.1 →  -6.0
-6.9 →  -6.0
-9.6  →  -9.0
-9.99x →  -9.0