fix(stm32): handle half-duplex in ringbuffered read

[badrbouslikhin]

This PR fixes an issue introduced in PR #3379, where the transmitter disable/receiver enable operation in half-duplex mode was moved from the flush method to the read methods. However, that PR did not account for the ringbuffered read method.

[Dirbaio]

@badrbouslikhin it seems this is causing the stm32l152re/usart_rx_ringbuffered test to fail

[badrbouslikhin]

I checked the reference manual (RM0038) and found no obvious reason for the test failure.
Also, the test does not appear to use a half-duplex USART, so the following condition should fail: if r.cr3().read().hdsel() && r.cr1().read().te() and this commit shouldn't be doing much.

I am not sure what's going on here, I'll check if I have a board stm32l1xxx board I can test at the office.

[badrbouslikhin]

I couldn't find a stm32l1xxx board. I can't debug this further at the moment.

Can you check it's not an issue with the physical setup of that board? I checked recent PRs, HIL tests didn't seem to run.

[Dirbaio]

okay I think the tes tis just flaky. It already randomly fails, this PR just makes it fail more often. Let's just ignore it.

Had to create a new PR #4054 . The next time please enable the "allow maintainers to push" checkbox when creating the PR.