add ExponentialBucketsRange function

[sbunce]

This function calculates exponential buckets with different arguments
than the existing ExponentialBuckets function. Instead of specifying the
start and factor, the user can specify the min and max bucket value. We
have been doing it this way internally at my company for some time.

[bwplotka]

I love it, thank you for your contribution!

[dswarbrick]

For the record, tests which expect bit-identical floating points will probably fail on s390x and possibly some other less mainstream archs, due to subtle differences in Go's floating point implementation on those archs, i.e. https://go.dev/ref/spec#Floating_point_operators, https://groups.google.com/g/golang-dev/c/Sti0bl2xUXQ

I'm seeing the exponential bucket range test fail on Debian's s390x buildds, and I suspect that it's for the same reason that this fails gohugoio/hugo#6387

=== RUN   TestBuckets
    histogram_test.go:364: exponential buckets range: got [1 1.6681005372000588 2.782559402207125 4.641588833612779 7.742636826811271 12.91549665014884 21.54434690031884 35.93813663804628 59.94842503189411 100.00000000000003],
                                                     want [1 1.6681005372000588 2.782559402207125 4.641588833612779 7.742636826811273 12.915496650148842 21.544346900318846 35.93813663804629 59.94842503189414 100.00000000000007]
--- FAIL: TestBuckets (0.00s)

[sbunce]

@dswarbrick
I did not know that! Thank you for educating me. I created a change which switches to a comparison with a tolerance "abs(a-b)<tolerance". The tolerance I picked "0.001" looks higher than the differences in your test failure output.

The PR is here.
#1133