Always send version(4) when resetting

[puddly]

This is the root cause of home-assistant/core#102503.

Firmware crashes were seen on EZSPv6 but not on the v8 stick I was testing when I initially added this special case. A HUSBZB-1 running EZSPv8, however, still crashes. This PR removes the protocol version limitation for removing this config.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (9cc1cde) 99.83% compared to head (fa56666) 99.83%.

❗ Current head fa56666 differs from pull request most recent head 7370b7b. Consider uploading reports for the commit 7370b7b to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##              dev     #589      +/-   ##
==========================================
- Coverage   99.83%   99.83%   -0.01%     
==========================================
  Files          68       68              
  Lines        4882     4881       -1     
==========================================
- Hits         4874     4873       -1     
  Misses          8        8              

Files	Coverage Δ
bellows/ezsp/__init__.py	99.40% <100.00%> (-0.01%)	⬇️

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[MattWestb]

I think EM35X and ÉFR32GM1X devices is getting short of RAM then stretching them to much but its also depends of the config of the firmware and its little different version around here.
MG12, MG13 and all MG2X shall not getting problems with RAM so fast then they is having plenty of it as long the firmware is OK configured.
I only running one EZSP / NCP with MG1P (Billy) and its restarting little now and then but no big problems (all test system is running RCP).

[puddly]

I'm renaming this PR because the problem is not what I initially suspected.

The root cause is...sending version(4) is required when you reset! Even though version(8) appears to work and the protocol version is upgraded, random crashes appear during startup. Sending version(4) before version(8) fixes the CONFIG_PACKET_BUFFER_COUNT issue as well.