Ensure preloading supports byte granularity

[creinwar]

Currently, HEX file lines that end between word boundaries are silently dropped.
This fix pads them with 0 to a full word to allow preloading.

[phsauter]

Wouldn't it make more sense to add something like the following after the inner while loop to 'clean up' remaining bytes?

if (byte_count != 0) begin
  data >>= 8 * (4 - byte_count); // zero-pad remaining bytes
  jtag_write(dm::SBData0, data);
  addr += 4;
  byte_count = 0;
  data = 32'h0;
end