[Profiling] Use TDCC group's `go` to infer first state's start time

[ayakayorihiro]

This PR contains small revisions to profiling:

1. We now show a summary for each group before showing the complete information.
2. We now use the corresponding TDCC group's go signal in order to infer when the FSM is technically "starting".

To describe a bit more about (2) - I previously had a bug where read from language-tutorial-iterate.futil was "active" for ~60 cycles (when it should only be active for 16). This was because read corresponds to the 0th state of the FSM of the par arm containing read, write, and upd, and the 0th state corresponds to both the first group and a reset state. To fix this, I used the go signal for the TDCC group that manages the FSM to figure out when the first group is active.