improve performance of _get_free_channel_id, fix channel max bug

[pawl]

This PR fixes a performance issue introduced by #377.

I noticed while running tests that _get_free_channel_id is way slower than I expected.

Surprisingly, this code that's similar to _get_free_channel_id (when all the channel ids are used) takes almost 40 seconds to run:

import time
from array import array

channel_max = 65535
used_channel_ids = array('H', range(1, channel_max + 1))
start = time.time()
for channel_id in range(1, channel_max + 1):
    if channel_id not in used_channel_ids:
        assert False
print(f'finished {time.time() - start:0.3f}')

This PR casts _used_channel_ids to a set before the loop, and it only takes ~0.004 seconds to run the loop after this change. It also allows keeping the memory usage improvements by keeping the _used_channel_ids stored as an array on the class.

I also learned that there's a bug preventing _get_free_channel_id from reaching the channel_max (the limit is currently channel_max - 1), and I fixed that too.

[auvipy]

can you also adjust some integration tests as well?

[pawl]

@auvipy Is this what you were thinking?: 3db6df1

[auvipy]

yes

[skinnyh]

Hi, can we get a version bump for this fix? Thank you!

[auvipy]

Yes of course, this weekend for sure

[skinnyh]

Yes of course, this weekend for sure

Hi, any update on it?

[GuardianRain]

Hi, which version has already include this fix? Thank you!

[auvipy]

5.1.1 please