[subset] code points after 0 in --unicodes argument ignored

[blueshade7]

It appears as if 0 is treated as a string null terminator.

% hb-subset --unicodes=41,42,0,43,44 --output-file=output.ttf Roboto-Regular.ttf
% spot -tcmap=5 output.ttf
--- encoding[0]
platformId=0
scriptId  =3
languageId=0
--- [code]=glyphId
[0041]=1 
[0042]=2 

--- encoding[1]
platformId=3
scriptId  =1
languageId=0
--- [code]=glyphId
[0041]=1 
[0042]=2 

--- encoding[2]
platformId=3
scriptId  =10
languageId=0
--- [code]=glyphId
[00000041]=1 
[00000042]=2 

In comparison, fontTools respects code points after 0.

% pyftsubset --unicodes=41,42,0,43,44 --output-file=output.ttf Roboto-Regular.ttf
% spot -tcmap=5 output.ttf
--- encoding[0]
platformId=0
scriptId  =3
languageId=0
--- [code]=glyphId
[0000]=1 
[0041]=2 
[0042]=3 
[0043]=4 
[0044]=5 

SNIP

[behdad]

It's a problem in how we represent text in utils/. I'll take a look.