Provide a stale_age to mkpidlock

Normally the pidlock should be held for a short duration; waiting for
its age to exceed the given limits of 10 seconds (for the fake gaproot,
which normally finishes in milliseconds) resp. 300 seconds (for
installing packages, which only takes a few seconds for most, but a few
bad ones can take minutes) until we consider it stale should be safe in
each use case. In fact Julia waits 5 times longer if the process
creating the pid lock file seems to be still running.

On the other hand, without a stale age, the lock file is *never*
considered stale, even if the process creating it definitely is gone,
and so the user can get stuck, which obviously is very bad. To
get unstuck they need to manually delete the lock file.

src/packages.jl

@@ -243,7 +243,9 @@ function install(spec::String, version::String = "";
     # point PackageManager to the given pkg dir
     Globals.PKGMAN_CustomPackageDir = GapObj(pkgdir)
     mkpath(pkgdir)
-    Pidfile.mkpidlock("$pkgdir.lock") do
+    # pidlock timeout: 300 seconds = 5 minutes should suffice; about the
+    # slowest packages to install are Semigroups and NormalizInterface.
+    Pidfile.mkpidlock("$pkgdir.lock"; stale_age=300) do
       if quiet || debug
         oldlevel = Wrappers.InfoLevel(Globals.InfoPackageManager)
         Wrappers.SetInfoLevel(Globals.InfoPackageManager, quiet ? 0 : 3)

src/setup.jl

@@ -159,7 +159,7 @@ function regenerate_gaproot()
 
     # create the mutable gaproot
     mkpath(gaproot_mutable)
-    Pidfile.mkpidlock("$gaproot_mutable.lock") do
+    Pidfile.mkpidlock("$gaproot_mutable.lock"; stale_age=10) do
         # create fake sysinfo.gap
         unquoted = Set(["GAParch", "GAP_ABI", "GAP_HPCGAP", "GAP_KERNEL_MAJOR_VERSION", "GAP_KERNEL_MINOR_VERSION", "GAP_OBJEXT"])
         open("$gaproot_mutable/sysinfo.gap", "w") do file