Partial network rewrite

[S-S-X]

TESTING REQUIRED

ANY OLD BUGS THAT ALSO APPEAR BEFORE THIS PR SHOULD NOT BE FIXED IN THIS BRANCH

Found new bugs so far:

• Supply converters are not working correctly, this seems to be because of machine timeout ABM change.
  actually they do seem to work but keeps blinking infotext between "no network" and actual supply info.
  Tests added ✔️
• Placing cable between other same tier cables and machine wont add machine to network (but other cables without network and machines after those will be added) Digging and placing machine again fixes connection.
  Tests added ✔️
• Digging/replacing connecting intermediate cable and then digging cable below machine wont remove machine from network, machine works and is part of network also drawing power normally from network.
  Tests added ✔️
• Digging machine will remove neighbor machines from network (but wont touch cables).
  Digging machines should not affect other machines at all, machine here behaves like cable and it should not.
  Tests added ✔️
• Placing machine adds neighbor unconnected cables to same network machine is added to.
  While placing machine should not affect unconnected cables it should still check for cables connected to another network and cause overload immediately, better sooner than later.
  This invalid network part is lost when cache is built again, only affects placement. Can be annoying for sure if you don't know how it behaves.
  Tests added ✔️
• Digging single cable will remove machine from network even if machine has multiple connected cables to same network
  Tests added ✔️
• Placing new cables around machine will always add machines to network, this can cause infinite duplicates in network.
  Tests added ✔️
• Machines wont time out ever when removed from network by digging cables, this causes infotext "has no network" to be immediately replaced with "idle" or "unpowered".
  No tests added ❌
• Networks wont overload immediately after placing machine between 2 different networks, not actually big problem but nice to have also for future updates (less aggressive overloads). This kind of worked before as caches were built a lot more often.
  Tests added ✔️
• Supply converter inserts duplicates to RE_nodes when cable connecting to network is placed nearby.
  No tests added ❌
• When supply converter is placed between clean networks it does not attach to networks correctly. This probably has something to do with cable tiers as when placed machines only look for first found compatible tier for cables.
  Tests added ✔️
• Digging supply converter does not correctly remove it from all networks, whole multi tier machine support is broken.
  Tests added ✔️

Just for the record actual changed LoC is not as bad as it looks like. Excluding test related stuff which causes 16 files changed, 2434 insertions(+), actual code changes are:
9 files changed, 982 insertions(+), 745 deletions(-)
Might be possible to optimize that bit but not sure if it is worth it here, there's other upcoming updates planned anyway.

#96 (comment)

Partial power network rewrite. Testing required, WIP, back off if you fear bugs

Opening PR for partially rewritten power network handler. Breaks compatibility if some other mod depends on SS metadata, however for special cases where it cannot be easily fixed on other mod it is very easy to add metadata back and update it with ABM.

Performance

Performance seems to be slightly better that what it was, with worst case network (built badly, multiple switching stations etc.) performance boost during testing (averages for more than 8k measurements) was over 50%. With "normal" networks not that much but still better. Didn't find any case where performance would be worse than before.

Help wanted

Most probably there will be bugs, for that reason I added label "Help wanted" hoping that someone would have time to test things and write/post screenshots about problems found. maybe even fix problems.

Highlights / most important stuff

Most important improvements:

• Switching station is not anymore network controller, actually it is not even registered as technic machine.
  It is still needed as it does add heart beat signal for network (resets network TTL countdown).
• Every switching station now displays supply / demand just like power monitor.
• Second ABM for switching stations removed.
• Globalstep executes networks directly instead of through switching station.
• Cache rebuilding is not triggered as often as it was before.
• Cache building is cheaper, flattening arrays removed by writing indexed arrays from beginning.
• More reusable code and is actually reused.
• Digging and building network nodes should be bit better and cheaper but no hard proofs on this.
• Added unit testing with busted along with some tests.

Removed metadata from switching station (moved values to network hash):

• active
• active_pos
• supply
• demand
• battery_count
• battery_charge
• battery_charge_max

Metadata cleanup

Added LBM (hardcoded disable) to remove metadata from technic machines, contains metadata keys for this PR and previous merged PR's:
"LV_EU_timeout", "MV_EU_timeout", "HV_EU_timeout", "LV_network", "MV_network", "HV_network", "active_pos", "supply", "demand", "battery_count", "battery_charge", "battery_charge_max", "active"

[S-S-X]

Things noted:
technic.cables is hash table containing all cables and all machines. This has not changed.
technic.networks[id].all_nodes was hash table containing network cables but not machines. This is updated to contain also machines.

At least naming has been wrong, added FIXME note to unit tests about technic.cables, it seems that many pieces assumes that machines are there so machines cannot be just simply dropped from technic.cables.

[S-S-X]

Cleaned up and merging to master, users test rest best in production right