std.enums: fix EnumSet.init and EnumMap.init for non-exhaustive enums

[sjb3d]

Proposed fix and EnumSet test case, fixes#19418.

[castholm]

The failing test is the std_enums_big_enums standalone test which verifies that enums with thousands of fields can participate in std.enums data structures without requiring overriding the comptime branch eval quota. Now that there is an additional call to Indexer.indexOf at comptime there will be more backward branches, so you will need to comb through the @setEvalBranchQuota expressions in std.enums and increment them to the new required bare minimum as appropriate.

As a starting point I would suggest iteratively running zig build-obj test/standalone/std_enums_big_enums.zig -fno-emit-bin and modifying the @setevalBranchQuota expressions until you find the new minimum quotas.

[sjb3d]

Ah thanks for the info. I hadn't realised that indexOf is a search when the enum is not dense. It looks like this will be n^2 cost for this case, so I'll rework the PR to only use indexOf for this non-exhaustive case, which should stay under the current limits. (I think!)

[sjb3d]

Latest in branch seems to pass locally, will need approval to trigger the CI workflow though. Thanks!