2F1 continuation recurrence can fail

[MikaelSlevinsky]

julia> HypergeometricFunctions._₂F₁continuation(-0.4, 0.0, 0.5, 0.5, 0.75+0.75im)
1.032 + 0.023999999999999997im

julia> HypergeometricFunctions._₂F₁continuation(-0.4, 100eps(), 0.5, 0.5, 0.75+0.75im)
1.030543999999993 + 0.019008000000025938im

This happens because the 3-term recurrence becomes effectively 2-term (but with a zero difference every second step) and the stopping criterion decides that two successive approximations are close because they're equal.