We have our own temporary file creation function now in
`git_futils_mktmp`, remove the others since they may be terrible on some
platforms.

`mktemp` on mingw is exceedingly deficient, using a single monotonically
increasing alphabetic character and the pid.  We need to use our own
random number generator for temporary filenames.

Introduce `git_rand`, a PRNG based on xoroshiro256**, a fast,
all-purpose pseudo-random number generator: https://prng.di.unimi.it

The PRNG will be seeded by the system's entropy store when possible,
falling back to current time and system data (pid, uptime, etc).
Inspiration for this was taken from libressl, but since our PRNG is
not used for cryptographic purposes (and indeed currently only generates
a unique temp file name that is written in a protected directory),
this should be more than sufficient.

Our implementation of xoroshiro256** was taken almost strictly from
the original author's sources, but was tested against PractRand to
ensure that there were no foolish mistranslations:

```
RNG_test using PractRand version 0.94
RNG = RNG_stdin64, seed = unknown
test set = core, folding = standard (64 bit)

rng=RNG_stdin64, seed=unknown
length= 256 megabytes (2^28 bytes), time= 2.9 seconds
  no anomalies in 210 test result(s)

rng=RNG_stdin64, seed=unknown
length= 512 megabytes (2^29 bytes), time= 6.2 seconds
  no anomalies in 226 test result(s)

rng=RNG_stdin64, seed=unknown
length= 1 gigabyte (2^30 bytes), time= 12.7 seconds
  no anomalies in 243 test result(s)

rng=RNG_stdin64, seed=unknown
length= 2 gigabytes (2^31 bytes), time= 25.4 seconds
  no anomalies in 261 test result(s)

rng=RNG_stdin64, seed=unknown
length= 4 gigabytes (2^32 bytes), time= 50.6 seconds
  no anomalies in 277 test result(s)

rng=RNG_stdin64, seed=unknown
length= 8 gigabytes (2^33 bytes), time= 104 seconds
  no anomalies in 294 test result(s)
```

push: Prepare pack before sending pack header.

odb: check for write failures

Co-authored-by: Edward Thomson <ethomson@github.com>

For large pushes, preparing the pack can take a while. Currently we
send the pack header first, followed by preparing the pack and then
finally sending the pack. Unfortunately github.com will terminate
a git-receive-pack command over http if it is idle for more than 10
seconds. This is easily exceeded for a large push, and so the push is
rejected with a Broken Pipe error.

This patch moves the pack preparation ahead of sending the pack header,
so that the timeout is avoided.

prepare_pack() can be called multiple times but will only do the work
once, so the original PREPARE_PACK call inside git_packbuilder_foreach()
remains.

SHA256: early preparation

revparse: support bare '@'

C90: add inline macro to xdiff and mbedtls

merge: comment conflicts lines in MERGE_MSG

merge: support zdiff3 conflict styles

ci: improve the name in CI runs

git has started adding comment markers to its conflict lines in
MERGE_MSG.  Match that behavior.

A bare '@' revision syntax represents HEAD.  Support it as such.

Disabling setting `CMAKE_FIND_LIBRARY_SUFFIXES` on Apple platforms.

refs: Speed up packed lookups.

Update to the xdiff used in git v2.35.0, with updates to our build
configuration to ignore the sort of warnings that we normally care
about (signed/unsigned mismatch, unused, etc.)

Any git-specific abstraction bits are now redefined for our use in
`git-xdiff.h`.  It is a (wildly optimistic) hope that we can use that
indirection layer to standardize on a shared xdiff implementation.

Deprecate the `git_packfile_hash` function.  Callers should use the new
`git_packfile_name` function which provides a unique packfile name.

The index's checksum is not an object ID, so we should not use the
`git_oid` type.  Use a byte array for checksum calculation and storage.

Deprecate the `git_indexer_hash` function.  Callers should use the new
`git_indexer_name` function which provides a unique packfile name.

The index's checksum is not an object ID, so we should not use the
`git_oid` type.  Use a byte array for checksum calculation and storage.

Deprecate the `git_index_checksum` function without a replacement.  This
is an abstraction that callers should not care about (and indeed do not
seem to be using).

Remove the unused `git_index__changed_relative_to` function.

A simple hash-to-hexadigit formatter.

The `git_oid_tostr_s` helper is indeed helpful, unless you are using
printf debugging (by inserting more `git_oid_tostr_s` calls) shortly
after using it.  Avoid it before invoking complex functions.

Underscores in filenames in tests indicate a logical separation - so
`commit_graph` becomes `commit::graph`.  Rename it to `commitgraph` to
avoid this.

COPYING: remove regex copyright, add PCRE copyright