enhance CANPacker with efficient MessageData struct

[deanlee]

This PR introduces a more efficient MessageData struct in the CANPacker class to significantly reduce map lookup overhead during CAN message packing. The original code performed excessive lookups, including inefficient searches using std::pair<address, string> keys. For example, the following code snippet includes five separate map lookups:

auto sig_it_counter = signal_lookup.find(std::make_pair(address, "COUNTER"));
 if (!counter_set && sig_it_counter != signal_lookup.end()) {
   const auto& sig = sig_it_counter->second;

   if (counters.find(address) == counters.end()) {
     counters[address] = 0;
   }
   set_value(ret, sig, counters[address]);
   counters[address] = (counters[address] + 1) % (1 << sig.size);
 }

By optimizing the way signals, counters, and checksums are accessed, the new structure improves performance and simplifies the code.

Key improvements:

• Introduced a MessageData struct for better organization and efficient management of signals, counters, and checksums.
• Kept pointers to signals in MessageData instead of copies, as these pointers are always valid in the dbc() instance—avoiding unnecessary data duplication.
• Minimized redundant lookups to significantly speed up the signal packing process.
• Cached counter and checksum signals directly in the MessageData struct to simplify counter and checksum handling.
• Optimized the pack method for faster processing of multiple signals, reducing overhead.

These enhancements lead to faster CAN message packing, particularly for larger datasets and frequent updates.