Remove absent messages

[edwardalee]

This is a companion to lf-lang/lingua-franca#2118. The main purpose of this PR is to remove absent messages from most federated programs, restricting their use to zero-delay cycles.

[byeonggiljun]

I tried to define and formularize what strict EIMT (SEIMT) should be and this is the result. I used the term SEIMT for simplification.

The RTI should send PTAG_i to federate i only when
i. EIMT_i is equal to the G(NET_i) from federate i.
ii. federate i is in a zero-delay cycle (ZDC).
iii. it can guarantee that there will be no messages that G(MSG_ij) <= G(PTAG_i) from upstream of i that are not in ZDC (j is a member of "immediate upstream federates that are not in a ZDC").

SEIMT is needed for checking the third condition. Now, let's examine cases of connections between federates.

1. A source federate is in a zero-delay cycle (ZDC).
   In the first case, the network message action of this connection in federate i will wait for a tagged message or a port absent message and everything will be ok. Please note that a source federate sends ABS messages when it is in a ZDC.
2. A source federate is not in a ZDC.
   SEIMT is important in this case. As a result of zero_delay_cycle_action_table, a federate does not wait for messages from its upstreams that are not in a ZDC. It just assumes the port is absent when no message has arrived.
   Thus, SEIMT_i should be calculated by min(EIMT_j + Delay_j) where Delay_j is a minimum delay between the federate j and the target federate i. When SEIMT_i > NET_i, the RTI knows that messages that G(MSG_ij) <= G(NET) will never be sent and is able to grant PTAG_i that G(PTAG_i) = G(NET_i).

With this declaration, SEIMT is calculated recursively as EIMT calculation uses recursion. However, I'm worried about the complexity of this calculation. Maybe we can save the result of update_min_delays_upstream somewhere and recalculate it only when the structure of the federation is changed?

@edwardalee @hokeun @lhstrh I would appreciate it if you read this and give me any feedback.

[edwardalee]

Yes, this looks right. update_min_delays_upstream already caches the results.

[lhstrh]

Subsumed by #323.