add lookup_user_record()

[vanosg]

Found by: Mystery-X
Patch by: Geo

It was discovered after the latest rewrite lookup up user records based on account that there are situations where m might not be known, and m being NULL can cause a crash if not properly checked. This creates a wrapper function to a) check which lookup to use based on the provided input, b) fallback to a different lookup in case something returns NULL.

Wrapper function to find an Eggdrop user record based on either a provided
channel memberlist record, host, or account. This function will first check
a provided memberlist and return the result. If no user record is found (or
the memberlist itself was NULL), this function will try again based on a
provided account, and then again on a provided host.

When calling this function it is best to provide all available independent
variables- ie, if you provide 'm' for the memberlist, don't provide
'm->account' for the account, use the independent source variable 'account'
if available. This allows redundant checking in case of unexpected NULLs

[thommey]

More generally I think I disagree with changing get_user_from_member(m) to lookup_user_record(m, NULL, NULL) or even lookup_user_record(m, NULL, from).
lookup_user_record(m, NULL, from) suggests m could be NULL when in almost all of those cases it can't be.
Using get_user_from_member(m) is less confusing in my opinion when we know m isn't NULL.

The nickname search combination lookup_user_record & check_all_chan_records should be folded into a single thing in my opinion, because it is always used together and is an unnecessary extra step. It could just get a nickname and call get_user_from_member(m) on the first one it finds and return that result.

[vanosg]

The (m, NULL, NULL) is necessary and only used in cases where host is derived from m. If m doesn't exist, m-> doesn't exist. These cases don't have a 'from' or similar variable to pull the host from.

[vanosg]

should be u2 = lookup_user_record(NULL, NULL, host); ?