fixing sec metadata

[john-friedman]

Old SEC submissions contain files in .txt form that are actual uuencoded pdfs or graphics. the filename is recorded by the sec as sequence + .txt.

for example 2.txt has content:

begin 644 10K.PDF
M)5!$1BTQ+C(-"B7BX\_3#0HQ(#`@;V)J#0H\/`T*+U!R;V1U8V5R("A&;&]W
M1&]C(#(N,"D-"B]+97EW;W)D<R`H4$1&("TM(&-Y8VQE(#`T*0T*+T-R96%T
M:6]N1&%T92`H1#HR,#`P,#(X,30U,#(X*0T*+TUO9$1A=&4@*$0Z,C`P,#`R
M,3,Q-#`T,#$I#0HO075T:&]R("A-97)R:6QL($-O<G!O<F%T:6]N*0T*+T-R
M96%T;W(@*$UE<G)I;&P@0V]R<&]R871I;VX@3F5T=V]R:R!#;VUP;W-I=&EO
...

this could be uudecoded using existing secsgml2 functions.