stream results when feasible

[BurntSushi]

Currently, all search results are written to an intermediate buffer in memory before being actually emitted to stdout. This is done to permit more efficient parallelism when searching. That is, only one thread can be writing to stdout at any point in time, but multiple threads can write to their own thread local memory buffer.

This does have undesirable end user consequences:

1. It can result in high memory usage when the number of search results is high.
2. When searching a single file, no output is seen until the search is complete.

We should be able to do quite a bit to fix these issues:

1. If a single file is given to search, then don't try any parallelism and make searching write to stdout directly.
2. If --threads 1 is given, then do (1) regardless of the number of inputs.

[BurntSushi]

This is done, but I opted to only do it for searching single files or stdin. Setting --threads 1 will still use the intermediate buffer.