Also record tests for the new behavior.

This is a semantic change, but more in line with how alists and
plists are defined in the CLHS.

See '15.6. Association Lists' and 'property list' in the Glossary.

Reported by Christoph Arenz to alexandria-devel in
'Wrong handling of duplicate keys in alist-hash-table and plist-hash-table'.

Move the definition of ensure-gethash at the top if the file
to prepare for the next change.

Patch by Bike from #lisp, http://paste.lisp.org/display/338772

fix typos in docstring and error messages.

Based on Kyle Littler's patch.

at the expense of an extra multiply. The unstable version can often be
converted to a fast instruction on a lot of machines, though this is
machine/implementation specific. As alexandria is more about correct code,
than efficiency, and we're only talking about a single extra multiply,
many would prefer the stable version.

Patch by github.com/mfiano.

Using the implementation's EMPTYP has the potential advantage of
supporting user-defined sequence types (as is the case with SBCL).

This commit tests for SEQUENCE:EMPTYP and defines ALEXANDRIA:EMPTYP in
terms of it.

For implementations that do not provide SEQUENCE:EMPTYP, nothing
changes.

EMPTYP usually (except, maybe, for some user-defined sequence types) has
two advantages:

1. it has O(1) runtime while LENGTH can be O(n)

2. it can work on user-defined, infinite sequences (given implementation
   support and integration of ALEXANDRIA:EMPTYP with the
   implementation's mechanism as is the case for SBCL)

  export and document them

  use them in READ-FILE-INTO-{STRING,BYTE-VECTOR}

  no smoke test for READ-STREAM-CONTENT-INTO-BYTE-VECTOR due to lack of
  something like WITH-INPUT-FROM-BYTE-VECTOR

Fixed problems:

1) Unintended variable capture affecting KEY, HASH-TABLE and DEFAULT
   FORMS
2) Multiple evaluation of the KEY and HASH-TABLE forms

ARRAY-DIMENSION-LIMIT is an exclusive upper bound.

Signed-off-by: Attila Lendvai <attila.lendvai@gmail.com>

REMF does not remove duplicate keys, and re-scanning
the list for each key is slower.

Signed-off-by: Attila Lendvai <attila.lendvai@gmail.com>

Previously, BISECT-BIG was never called.

Suggested by Stas Boukarev.

Patch by "James M. Lawrence" <llmjjmll@gmail.com>

Signed-off-by: Attila Lendvai <attila.lendvai@gmail.com>

  add IO chapter containing READ-FILE-INTO-{STRING,BYTE-VECTOR}

  change include/fun-alexandria-ensure-gethash.texinfo
  -> include/macro-alexandria-ensure-gethash.texinfo

 Return NIL if the sequence is empty, instead of the NO-EXTREMUM nonsense.

 It was bad design, because it's not an error someone higher up the stack can
 sensibly handle, and handling it locally is too verbose and slow.

    (or (extremum ...) (error ...))

 expresses the common case succintly, and fits the pattern of existing
 sequence functions.

 If it is deemed necessary, we can also add &KEY DEFAULT, but that seems
 overkill and has little precedent in sequence functions.

  From http://www.cliki.net/EXTREMUM

  A simple version built on top of REDUCE for now.

  Yech. Phui. Tabs.

  Workarounds for the Clisp and CMUCL failures.

  Newer Clisp's no longer have the stack-overflow for the hash-table test, so
  enable it too.