Inaccuracy observed in bairy function

[jacksonvanover]

The jmat implementation of the bairy function appears to exhibit inaccuracies around the point -9.5382153125 when compared against the mpmath and gsl implementations. (The input was discovered via s3fp)

evaluating bairy(-9.53):

• jmat evaluates to 0.008119067602717238
• mpmath and gsl evaluate to 0.00812456792635

evaluating bairy(-9.538215):

• jmat evaluates to -0.000026382136231793024
• mpmath and gsl evaluate to -0.00002044715920

evaluating bairy(-9.5382153125):

• jmat evaluates to -0.000026692305447213503
• mpmath and gsl evaluate to -0.00002075702995

evaluating bairy(-9.54):

• jmat evaluates to -0.0017964554210120178
• mpmath and gsl evaluate to -0.00179041955022