Removed redundant strip of trailing zeros for 2.0 messages.

[beforezod]

Given the four helper functions: ..pack, ..to_send_buffer, ..parse_char, ..decode
both ...pack and ...to_send_buffer call the _mav_trim_payload function to trim the trailing zeros. Only the ..to_send_buffer call should do this. Otherwise you cannot pack a message, use the getter functions, and expect the same data back.

We ran into this problem because all our unit tests were attempting exactly this and began to fail when we switched to the v2.0 standard.

[hamishwillee]

Slightly related to #241 (ie in that both are related to zero trimming)

[hamishwillee]

This is failing with some parse errors

[amilcarlucas]

Can you please rebase this ? Or even better, cherry pick
01d6a75 on top of master

[beforezod]

Yes. Will do.

[beforezod]

This fail is interesting. It passes with my message set locally.

[beforezod]

GOPRO_HEARTBEAT { status: 5 capture_mode: 72 flags: 139 }
GOPRO_HEARTBEAT { status: 5 capture_mode: 72 flags: 139 }
GOPRO_GET_REQUEST { target_system: 5 target_component: 72 cmd_id: 139 }
GOPRO_GET_REQUEST { target_system: 5 target_component: 72 cmd_id: 139 }
GOPRO_GET_RESPONSE { cmd_id: 5 status: 72 value: [ 139, 140, 141, 142, ] }
GOPRO_GET_RESPONSE { cmd_id: 5 status: 72 value: [ 139, 140, 141, 142, ] }
GOPRO_SET_REQUEST { target_system: 5 target_component: 72 cmd_id: 139 value: [ 206, 207, 208, 209, ] }
GOPRO_SET_REQUEST { target_system: 5 target_component: 72 cmd_id: 139 value: [ 206, 207, 208, 209, ] }
GOPRO_SET_RESPONSE { cmd_id: 5 status: 72 }
GOPRO_SET_RESPONSE { cmd_id: 5 status: 72 }
RPM { rpm1: 17.000000 rpm2: 45.000000 }
RPM { rpm1: 17.000000 rpm2: 45.000000 }
Received 604 messages on channel 0 OK
Received 604 messages on channel 1 OK
No errors detected