Do not adjust delay if retry_after < delay

[bahaa32]

As it stands, it appears that the delete delay could be lowered when that may have not been the intention (the comment on the line says // increase delay). This PR adds a check before adjusting the delete delay such that it will only ever increase.

Possible issues:

• If the user starts the deletion process and leaves it unattended, a long retry_after response could unexpectedly cripple the deletion speed without the possibility of the delay decreasing to a more sane value.

Perhaps it is time to algorithmically discover the rate limit? #386