Wrap hackney:close in try/catch to prevent crash when ETS table is gone

[p424p424]

When the streaming process calls hackney:close/1 during cleanup, hackney's connection-manager ETS table may already have been torn down by another process, causing a bad argument crash in ets:lookup/2.

This is a race condition — it happens intermittently, always after the stream has finished and the response has been delivered, so it's cosmetic (exit code is still 0). But it produces a scary-looking CRASH REPORT in the terminal.

The fix wraps the close call in a try/catch, matching the pattern already used throughout hackney itself for similar cleanup operations.

Before

close(ClientRef) ->
  hackney:close(ClientRef),
  nil.

After

close(ClientRef) ->
  try hackney:close(ClientRef) catch _:_ -> ok end,
  nil.