Fix errors in hivemind.p2p and hivemind.compression

[borzunov]

This PR:

1. Fixes warnings in hivemind.p2p destructors:

   Exception ignored in: <function P2P.__del__ at 0x7fbec8be5f70>
   Traceback (most recent call last):
     File "/home/jheuristic/anaconda3/envs/py38_petals_borzunov/lib/python3.8/site-packages/hivemind/p2p/p2p_daemon.py", line 636, in __del__
       self._terminate()
     File "/home/jheuristic/anaconda3/envs/py38_petals_borzunov/lib/python3.8/site-packages/hivemind/p2p/p2p_daemon.py", line 657, in _terminate
       self._child.terminate()
     File "/home/jheuristic/anaconda3/envs/py38_petals_borzunov/lib/python3.8/asyncio/subprocess.py", line 141, in terminate
       self._transport.terminate()
     File "uvloop/handles/process.pyx", line 636, in uvloop.loop.UVProcessTransport.terminate
     File "uvloop/handles/process.pyx", line 378, in uvloop.loop.UVProcessTransport._check_proc
   ProcessLookupError:
   Exception ignored in: <function Client.__del__ at 0x7fbec8bd7a60>
   Traceback (most recent call last):
     File "/home/jheuristic/anaconda3/envs/py38_petals_borzunov/lib/python3.8/site-packages/hivemind/p2p/p2p_daemon_bindings/p2pclient.py", line 53, in __del__
       self.close()
     File "/home/jheuristic/anaconda3/envs/py38_petals_borzunov/lib/python3.8/site-packages/hivemind/p2p/p2p_daemon_bindings/p2pclient.py", line 50, in close
       self.control.close()
   AttributeError: 'NoneType' object has no attribute 'close'

2. Makes bfloat16 serialization in hivemind.compression forward- and backward-compatible. The code before this PR (a) didn't work in torch < 1.13.0 (hivemind requires torch >= 1.9.0) and (b) led to this warnings on torch >= 2.0:

   /opt/conda/lib/python3.10/site-packages/hivemind/compression/base.py:115: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly.  To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()
     tensor = torch.as_tensor(storage, dtype=torch.bfloat16)

   The new code works without warnings in all versions of PyTorch.

[codecov]

Codecov Report

Merging #565 (07e98ae) into master (6c3a46c) will increase coverage by 0.07%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #565      +/-   ##
==========================================
+ Coverage   75.85%   75.92%   +0.07%     
==========================================
  Files          81       81              
  Lines        8002     8001       -1     
==========================================
+ Hits         6070     6075       +5     
+ Misses       1932     1926       -6     

Impacted Files	Coverage Δ
hivemind/compression/base.py	94.28% <100.00%> (-0.24%)	⬇️
hivemind/p2p/p2p_daemon.py	94.47% <100.00%> (+0.01%)	⬆️
hivemind/p2p/p2p_daemon_bindings/p2pclient.py	100.00% <100.00%> (ø)

... and 2 files with indirect coverage changes

[justheuristic]

lgtm