fix: prevent latch inference

[tianrui-wei]

This PR prevents latch inference by assigning a default value to msg_send_data_size_S4

[tianrui-wei]

cc @Jbalkind @fei-g
The latch inference seems to come from a39066a. Would you have any insight on why the comments was inserted?

[fei-g]

I don't remember exactly why did I comment that out.. Just checked the L15 decoder, and I found that it didn't look into the data_size field. I think in L15 pipeline it will use the data size stored in MSHR. So, we are okay to set the default data_size to 0. But If you have other modules that need the data_size info, the best way might be just uncomment those lines back (or set default data size to data_size_S4_f).