sed script might be what you need:
sed -n '$p;N;s/[[:blank:]]*\n[[:blank:]]*;/;/;P;D' csv_file
Broken down for explanation:
$p; # at last line of stream, just print it
N; # append the next line from input so that we always consider two lines at a time ...
s/[[:blank:]]*\n[[:blank:]]*;/;/; # then replace `\n;` (and any leading and trailing blanks) with just `;`, and ...
P; # print only the _first_ of the two lines present in memory and ...
D # then delete it and read one new line if memory becomes empty
Note that the first command,
$p, actually prints something only when total number of input lines is odd, because the rest of the script ensures that there are always two lines from the input stream in memory