Wrong hashes for some Rationals

[sostock]

On Julia v1.10.0-alpha1 and master, some Rationals which are equal have different hashes:

julia> hash(5//2) # same hash as on older Julia versions (≥ v1.6)
0x16b7f40fdbb2c57d

julia> hash(Int128(5)//2) # the same hash is also obtained for UInt128(5)//2 and BigInt(5)//2
0xb795033f6f2a0674

[oscardssmith]

caused by #50041. The change in hash is intentional but the inconsistency isn't. There are a few bugs that were added that I just found. Thanks for the test case!

[oscardssmith]

My initial guess was partially wrong. The actual problem was caused by both #50041 missing hashing odd denominators combined with the fact that #49996 was incorrectly computing the position of the right most bit. PR incoming.