Lambert W Function (hp-42s)

[Albert Chan]

Hi lyuka:

I was wrong to suggest removing x=-1/e test. e^W(-1/e) take many tries to converge.
Because of catastrophically cancelled slope, we got something like this:

0.367879441171442
0.25
0.3051544444752
0.335540374899958
0.351461975177244
0.359608250433447
0.363728172008137
...

I would recommend reverting to previous version (with x=-1/e test)
Besides, previous version is much easier to understand.