Fix infer issues

[philljj]

Description

Fix UNINITIALIZED_VALUE bugs from infer testing.

Testing

Confirmed fixed with Infer rerun.

[philljj]

Retest this please.

[philljj]

Retest this please

[douzzer]

Looks good -- just a few overheady analyzer coddlings to redo in a non-costly way.

Passes all the latest analyzers locally in the wolfssl-multi-test.sh ... super-quick-check suite, which is good news -- it's entirely possible to coddle one analyzer only to have a different analyzer report an unused value or other defect around the coddling.

[douzzer]

This XMEMSET() incurs non-negligible overhead to address a clearly-false positive from infer. I can see how the analyzer got confused, given that wolfSSL_CTX_set_groups() is in another file (tls13.c), but we should try to find a better way to address this.

[douzzer]

This could've stayed on one line, à la

int ret = 0, row = 0;

(just a FWIW)

[douzzer]

As in wolfSSL_CTX_set1_groups(). this is an overheady way to address a clearly-false positive.

Also, any idea why infer didn't similarly false-positive on wolfSSL_set1_groups()? Weird.

[douzzer]

This XMEMSET() seems OK to me. It's not even immediately clear that it's a false positive, but mainly this is a rendering routine and the memset involves only 16 bytes.

[douzzer]

This is another expensive memset to address a clearly-false positive, as above. in wolfSSL_CTX_set1_groups() etc.

[douzzer]

Same story, expensive op to address a false positive.

[douzzer]

This addresses a false positive, but I can live with it, given it's only 6 bytes. Btw I probably would've resolved it by refactoring that neighborhood to convert the ASCII port to a word32 without the unnecessary copying of the ASCII.

[philljj]

Rebased and removed the added XMEMSET(groups and XMEMSET(_groups. We are considering the groups to be initialized by the for loop logic and so these are false positives.

[douzzer]

LGTM