Memory leak on TLS handshake

[maxtropets]

Based on Teams community memory leak post.

TLS handshake leaks memory after the update to the most recent symcrypt (surpassingly, v 1.8.0).

Somewhat similar to #109, but this time with a stable repro.

Repro can be found here: microsoft/CCF#6994.

UPD. Seems like it's not only handshake that leaks, more interactions with the session cause more same-size -ish leaks.

[maxtropets]

2025-05-07T00:34:14.0728404Z SUMMARY: AddressSanitizer: 1140 byte(s) leaked in 95 allocation(s).
2025-05-07T00:34:14.0939872Z ==2340==ERROR: LeakSanitizer: detected memory leaks
2025-05-07T00:34:14.0940083Z Direct leak of 6888 byte(s) in 574 object(s) allocated from:
2025-05-07T00:34:14.0940652Z SUMMARY: AddressSanitizer: 6888 byte(s) leaked in 574 allocation(s).
2025-05-07T00:34:14.0943523Z ==2340==ERROR: LeakSanitizer: detected memory leaks
2025-05-07T00:34:14.0943724Z Direct leak of 6888 byte(s) in 574 object(s) allocated from:
2025-05-07T00:34:14.0944294Z SUMMARY: AddressSanitizer: 6888 byte(s) leaked in 574 allocation(s).
2025-05-07T00:34:14.2023134Z ==2316==ERROR: LeakSanitizer: detected memory leaks
2025-05-07T00:34:14.2023287Z Direct leak of 3384 byte(s) in 282 object(s) allocated from:
2025-05-07T00:34:14.2023750Z SUMMARY: AddressSanitizer: 3384 byte(s) leaked in 282 allocation(s).
2025-05-07T00:34:14.2206421Z ==2334==ERROR: LeakSanitizer: detected memory leaks

Noticed in e2e test logs, it's presumably 12 bytes always

[mamckee]

Created a PR with the fix. This was a regression introduced in #112. Investigating why this was missed in our pipelines.

[mamckee]

Fixed by #116. Fix is tagged in the feature branch and pending integration into AZL3.