Configure split_words for all env vars about envconfig HOT 1 CLOSED

I think this complicates the API too much. We already have multiple ways to accomplish the same thing: envconfig:"MONGO_POOL_SIZE" vs split_words:"true". Mixing Process-level configuration with the field-wise configs we already have introduces a lot more potential for confusion. I assume the intention is that field tags would still override the global configuration? What about a global split_words mixed with envconfig field tags? That's nonsensical. As it is, one can look at the struct definition and see everything they need to know about its deserialization, just like the norm with other serdes libraries (json, yaml, etc).

from envconfig.