integrate(exp(-a*x)*exp(-b*y), (x, 0, oo), (y, 0, oo)) causes an maximum recursion depth error

[ghost]

>>> integrate(exp(-a*x)*exp(-b*y), (x, 0, oo), (y, 0, oo))
RecursionError: maximum recursion depth exceeded in comparison
>>> integrate(exp(-a*x), (x, 0, oo))
⎧    1                                       π
⎪    ─       for │periodic_argument(a, ∞)│ < ─
⎪    a                                       2
⎪                                             
⎪∞                                            
⎨⌠                                            
⎪⎮  -a⋅x                                      
⎪⎮ ℯ     dx              otherwise            
⎪⌡                                            
⎪0                                            
⎩                                         

In Latex printing, "periodic_argument" makes the word "argument" the subscript of the word "periodic"
I don't know what "periodic_argument(a, ∞)" means, should it be printed
I think that only the 1/a should be returned, together with the condition on a, I don't think the "otherwise" thing should be returned

>>> Sum(a**n*b**m, (m, 0, oo), (n, 0, oo)).doit()
⎛⎧   1                 ⎞ ⎛⎧   1                 ⎞
⎜⎪ ──────   for │a│ < 1⎟ ⎜⎪ ──────   for │b│ < 1⎟
⎜⎪ -a + 1              ⎟ ⎜⎪ -b + 1              ⎟
⎜⎪                     ⎟ ⎜⎪                     ⎟
⎜⎪  ∞                  ⎟ ⎜⎪  ∞                  ⎟
⎜⎪ ___                 ⎟ ⎜⎪ ___                 ⎟
⎜⎨ ╲                   ⎟⋅⎜⎨ ╲                   ⎟
⎜⎪  ╲    n             ⎟ ⎜⎪  ╲    m             ⎟
⎜⎪  ╱   a    otherwise ⎟ ⎜⎪  ╱   b    otherwise ⎟
⎜⎪ ╱                   ⎟ ⎜⎪ ╱                   ⎟
⎜⎪ ‾‾‾                 ⎟ ⎜⎪ ‾‾‾                 ⎟
⎜⎪n = 0                ⎟ ⎜⎪m = 0                ⎟
⎝⎩                     ⎠ ⎝⎩                     ⎠

I think it should return

        1        
─────────────────        for |a| < 1 and |b| < 1
(-a + 1)⋅(-b + 1)

I'm not sure it's worthwhile giving (or attempting to give) information other than this

[cbm755]

The single integral result is correct, because a could be a complex value.

Pretty printing looks better with development branch, it looks like the periodic_arg thing was improved by commit f18ce44

⎧    1           π           
⎪    ─       for ─ > │arg(a)│
⎪    a           2           
⎪                            
⎪∞                           
⎨⌠                           
⎪⎮  -a⋅x                     
⎪⎮ ℯ     dx     otherwise    
⎪⌡                           
⎪0                           
⎩  

(maybe that commit could have had a test).

[cbm755]

The case where a is real could be improved: see #14606.