Allow custom MSB setting in generated primes

[fjarri]

In the RSA application one needs two primes which, when multiplied, produce a number with a fixed bit length. This can be achieved by requiring two most significant bits to be set on each. The current code just sets the MSB by default, and setting the second one requires writing a custom wrapper over SmallPrimesSieveFactory.

This PR attempts to provide a simpler pathway for most common bit-setting cases: no bits set (primes generated up to the chosen bit length), MSB set, and two MSBs set. random_odd_uint() and SmallPrimesSieveFactory constructors get an additional SetBits parameter.

An RSA application would have to use the new functionality by calling

    sieve_and_find(
        rng,
        SmallPrimesSieveFactory::new(bit_length, SetBits::TwoMsb),
        is_prime_with_rng,
    )

The default currently remains SetBits::Msb, but perhaps it should be changed to SetBits::None.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.46%. Comparing base (838c157) to head (9dfe54d).
Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #69      +/-   ##
==========================================
+ Coverage   99.44%   99.46%   +0.01%     
==========================================
  Files          10       10              
  Lines        1449     1491      +42     
==========================================
+ Hits         1441     1483      +42     
  Misses          8        8              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dvdplm]

lgtm so far

[dvdplm]

LGTM.

One question I have is how this jives with #62. In my head this PR can go in straight away and we should not let #62 be a blocker (it will break semver and require a 0.7 release anyway).

[fjarri]

Sure, we can get this in first.