#5211 With radexpand:all, trigrat returns minus its own argument

[David Scherfgen]

radexpand: all$
trigrat(sqrt(sin(x)));    /* -sqrt(sin(x)) */
trigrat(sin(x)^(3/2));    /* -sin(x)^(3/2) */
trigrat(1/sqrt(sin(x)));  /* -(1/sqrt(sin(x))) */

Systematic for half-integer powers of a sine — also sqrt(sin(2*x)) and sqrt(sin(x))*sqrt(cos(x)) — and not for cosine, not for a cube root. Under exponentialize a sine carries a factor of -%i that a cosine does not, and with radexpand:all that factor is pulled out of the radical and comes back on the other branch.