Type= test fixes

See merge request alexandria/alexandria!37

Fix the modal logic of TYPE=.

See merge request alexandria/alexandria!29

test: fix tests on SBCL

Closes #38

See merge request alexandria/alexandria!33

Mark CURRY.3 test as XFAIL on CMUCL.

See merge request !30

Fix issue 30 - testing value of `hash-table-size`.

See merge request alexandria/alexandria!28

For whatever reason, CMUCL is unable to retrieve the value of
COMPILER-MACRO-FUNCTION for CURRY after the DEFINE-COMPILER-MACRO
call. I believe that this is a violation of the spec.

At any rate, to get the tests to pass on CMUCL, I have marked this as
XFAIL on CMUCL.

Computing the correct return value of TYPE= requires combining the
primary and secondary returns of SUBTYPEP correctly.  Previously the
return value for combining (NIL NIL) and (T T) was incorrectly (NIL T)
instead of (NIL NIL).

As far as I can tell from the spec, the meaning of `hash-table-size`
is not dictated, and implementations are free to return just about
anything; it's just a way to provide useful information (hints) when
creating a new hash-table.

I disabled this test on Allegro originally, but Attila Lendvai
proposes we simply remove it altogether, as this commit does.

fix some tests

See merge request alexandria/alexandria!27

Clasp image is broken.

1. Accept any error from `read-stream-content-into-byte-vector.2`

2. Temporarily disable `test=.3` on Allegro because of a bug in
Allegro's `subtypep` (RT has no XFAIL).

A current Allegro beta gives different values for these tests that
seem acceptable according to the CL ANSI Spec.

Specifically:

- the `hash-table-size` of a hash-table copy is not always equal to
the size of the original hash-table, but this seems to be permitted by
the spec. Gitlab issue #30.

- Allegro gives a `program-error` in one case where the tests expect a
`type-error`. GitLab issue #31.

- Allegro is more sure about a `subtypep` query than some other lisps
in a way that seems permissible.  GitLab issue #32.

Thanks for reporting, Robert!

Fix READ-STREAM-CONTENT-INTO-STRING on abcl and cmucl

See merge request alexandria/alexandria!24

Use modern ASDF style and define all systems in alexandria.asd

See merge request alexandria/alexandria!26

Add Gitlab CI

See merge request alexandria/alexandria!25

Paper over implementation incompatibilities in WITH-OPEN-FILE*.

Be smarter about element types

Closes #25

See merge request !23

The element type of the buffer should also be the element type of the
stream, rather than character.

This is a fix to the previous commit.

Default the element type of the string being created from the element
type of the stream being read.

This has two advantages: it will make things work for implementations
which have a mode where the default string element type is a subtype
of character, and it will also mean that if the stream's element type
is a subtype of character you may get thinner strings.

(From Doug)
    For context, I'm trying to eradicate all such problems (using undefined
    behavior sanitizer <https://sourceforge.net/p/sbcl/sbcl/ci/d931c7eb>) so
    that, among other things, I can figure out whether the non-zero-
    initialized DX array patch can ever be released.

Add functions relating to multidimensional array indexing conversions

See merge request alexandria/alexandria!20

Use linear-time SHUFFLE if supplied with many list items to shuffle

Closes #21

See merge request alexandria/alexandria!21

A purely cosmetic issue.