while here also wrap fcgi_send_response to make sure we don't end up trying to send a fcgi record bigger than allowed. ok tracey@

Rather than realloc in fixed-sized blocks, use the 1.5 * allocated scheme when growing the array. This produces fewer allocations and up to 3x speedup on large diffs. ok stsp@

Available to all views; for example, 99g will navigate to line 99 in the file if in a blame view or the 99th commit if in a log view. Includes improvements suggested by naddy and stsp. ok naddy@

ok tracey

ok op@

as originally intended by tracey@

fcgi_send_response can end up trying to send big amount of data, in particular when serving blobs. if a write failure occurs, gotwebd and httpd go out of sync and "bad stuff happens". debugged with and ok tracey@

there's no need to distinguish between plain text blob and a binary one in the output, just send it chunk by chunk. ok tracey@

ok tracey@

ok tracey@

suggested by stsp@

it's useful for e.g. to apply old patches that weren't generated by `got diff' or git-diff(1) and so lacks the metadata needed for the 3-way merge. manpage improvements from and ok stsp@

Current code is overly complex; it looks if the diff *might* use diff3 merge to postpone reverse_patch until we know for sure. Instead, just reverse_patch in apply_patch once we know if merge *is* used or not.

ok stsp@

`got patch' cannot read patches from non-regular files for obvious reasons. However, it could crash in sendmsg because pledge doesn't allow to send file descriptors referring to directories. So, restrict `got patch' to operate on regular files only and fail otherwise. This still allows to read patches from symlinks since they're resolved at open(2) time and the file type check is performed after. There may be a marginal usefullness in reading patches from fifos, but the current code doesn't allow that anyway since got-read-patch needs a seekable file descriptor anyway. ok tracey@