Fix dconst in templateTable_riscv32.cpp

* The current instruction only works with single-precision 0, so modified to load double-precision 0 from memory.