Removed "Meta" field from syn_registry_entry (syn_registry_by_name table) to reduce RAM usage

[emikodes]

Spawning a child process currently requires the insertion of two tuples, respectively in the syn_registry_by name and syn_registry_by_pid ETS.

In the current implementation, the Meta field is fully duplicated in both tables:

-type syn_registry_entry() :: {
    Name :: term(),
    Pid :: pid(),
    **Meta :: term(),**
    Time :: integer(),
    MRef :: undefined | reference(),
    Node :: node()
}.

-type syn_registry_entry_by_pid() :: { %%kept only here (First element of tuple becomes a key in ETS)
    Pid :: pid(),
    Name :: term(),
    **Meta :: term(),**
    Time :: integer(),
    MRef :: undefined | reference(),
    Node :: node()
}.

If Meta is a large piece of information, this duplication causes unnecessary memory usage.

In this PR, the syn_registry_by_name table has been shrinked by removing the Meta field.

Read operations now require a double-jump lookup:
1.Find the Pid from the Name (in syn_registry_by_name).
2.Find the Meta from the Pid (in syn_registry_by_pid).

This optimization allows to save up to 50% of the ETS Usage. Also, Write and Delete operations become marginally lighter due to less memory allocation and copying.

Executing a little "StressTest" script, which spawns 500k processes and shows the memory footprint, allows to visualize the memory reduction:

Before:

iex(2)> StressTest.run()
Spawning 500000 processes.

Initial BEAM memory:
   - Name Table: 0.0 MB
   - PID Table:  0.0 MB
   - ETS:   0.93 MB


Spawn completed
Elapsed time: 9.188 seconds
   - Name Table: 640.81 MB
   - PID Table:  640.81 MB
   - ETS:   1289.41 MB

After:

iex(3)> StressTest.run()
Spawning 500000 processes.
Initial BEAM memory:
   - Name Table: 0.0 MB
   - PID Table:  0.0 MB
   - ETS:   0.98 MB


Spawn completed
Elapsed time: 10.008 seconds
   - Name Table: 68.6 MB
   - PID Table:  640.81 MB
   - ETS:   717.21 MB